Georgia State University Georgia State University
ScholarWorks @ Georgia State University ScholarWorks @ Georgia State University
Economics Dissertations
10-30-2009
A Comparative Study of Fiscal Decentralization in China and India A Comparative Study of Fiscal Decentralization in China and India
Yinghua Jin Georgia State University
Follow this and additional works at: https://scholarworks.gsu.edu/econ_diss
Part of the Economics Commons
Recommended Citation Recommended Citation Jin, Yinghua, "A Comparative Study of Fiscal Decentralization in China and India." Dissertation, Georgia State University, 2009. https://scholarworks.gsu.edu/econ_diss/57
This Dissertation is brought to you for free and open access by ScholarWorks @ Georgia State University. It has been accepted for inclusion in Economics Dissertations by an authorized administrator of ScholarWorks @ Georgia State University. For more information, please contact [email protected].
PERMISSION TO BORROW
In presenting this dissertation as a partial fulfillment of the requirements for an advanced degree from Georgia State University, I agree that the Library of the University shall make it available for inspection and circulation in accordance with its regulations governing materials of this type. I agree that permission to quote from, to copy from, or to publish this dissertation may be granted by the author or, in his or her absence, the professor under whose direction it was written or, in his or her absence, by the Dean of the Andrew Young School of Policy Studies. Such quoting, copying or publishing must be solely for scholarly purposes and must not involve potential financial gain. It is understood that any copying from or publication of this dissertation which involves potential gain will not be allowed without written permission of the author.
_______________________________________ Signature of the Author
NOTICE TO BORROWERS
All dissertations deposited in the Georgia State University Library must be used only in accordance with the stipulations prescribed by the author in the preceding statement.
The author of this dissertation is: Yinghua Jin 1322 Briarwood Rd, H-10 Atlanta, GA, 30319 The director of this dissertation is: Dr. Mark W. Rider Andrew Young School of Policy Studies Georgia State University P.O. Box 3992 Atlanta, GA, 30302-3992
Users of this dissertation not regularly enrolled as students at Georgia State University are required to attest acceptance of the preceding stipulations by signing below. Libraries borrowing this dissertation for the use of their patrons are required to see that each user records here the information requested.
Type of use Name of User Address Date (Examination only or copying)
A COMPARATIVE STUDY OF FISCAL DECENTRALIZATION IN CHINA AND INDIA
BY
YINGHUA JIN
A Dissertation Submitted in Partial Fulfillment of the Requirements for the Degree
of Doctor of Philosophy
in the Andrew Young School of Policy Studies
of Georgia State University
GEORGIA STATE UNIVERSITY 2009
Copyright by Yinghua Jin
2009
ACCEPTANCE
This dissertation was prepared under the direction of the candidate’s Dissertation Committee. It has been approved and accepted by all members of that committee, and it has been accepted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Economics in the Andrew Young School of Policy Studies of Georgia State University.
Dissertation Chair: Dr. Mark W. Rider
Committee: Dr. Roy W. Bahl Dr. Shiferaw Gurmu Dr. Penelope B. Prime Dr. Yongsheng Xu
Electronic Version Approved:
W. Bartley Hildreth, Dean Andrew Young School of Policy Studies Georgia State University December 2009
vi
ACKNOWLEDGMENTS
I am deeply indebted to my chair, Dr. Mark W. Rider, who accepted me when I was in
the most difficult time of my life. His commitment helped drive me to make this finally happen. I
appreciate his precious advice and friendly reminders. This study would never have been
accomplished without the help of his profound knowledge, brilliant ideas, and sophisticated
methods. I also appreciate his allocation of weekends to polish my essays and his efforts in
seeking funding for me. I am also grateful for his generosity, including the use of his SkyMiles
to get an airline ticket to visit my family.
I am also indebted to my committee members, Dr. Roy W. Bahl, Dr. Shiferaw Gurmu, Dr.
Penelope B. Prime, and Dr. Yongsheng Xu. With their insightful comments, I was able to take a
step back from the individual trees and have another look at the forest. I would also like to thank
Dr. Paul Kagundu and Dr. Baoyun Qiao for their valuable comments at the proposal stage.
Special thanks go to Dr. Baoyun Qiao for sharing his data with me for the robustness check.
My gratitude also goes to many friends: Dr. Li Zhang, Dr. Weimin Zhang, Dr. Matthew
Sartin, Dr. Wenchun Zhang, Dr. Zhiqiang Yan, Ping Meng, Yongzheng Liu, Xi Rao, Junwei Zou,
Guchuan Yang, Wenhui Huo, Huiping Du, Abdullah Khan, Harini L. Kannan, and other fellow
graduate students who began in 2004, for their data support, draft editing, document printing, and
numerous encouragements to overcome my depression and autism.
My deep appreciation and respect go to Judge Robert E. Flournoy, III; his wife, Julie B.
Flournoy; his mother, Pam Flournoy; and other family members for their great kindness and help
throughout my American life. Judge Flournoy and his family provided me with a valuable
internship opportunity, introduced me to the common law system, sponsored a museum
vii
membership, and invited my family and me to spend many holidays together with them. I and
my family have enjoyed staying with them so much.
My appreciation also goes to Hanxu Yu and Mark S. Gatewood for their part-time work
opportunities and personal help.
My thanks also go to many other Chinese and American friends in China and in the
United States. I will not name them here but I did enjoy all the get-together time with them.
Thank you, my friends, brothers and sisters.
My deepest love, gratitude, and indebtedness go to Dad, Damin Jin, who rests in peace in
another world; Mom, Xingchun Luo, who still lives alone in a remote mountainous area waiting
for me; my elder brother, Yingde Jin; and two elder sisters, Yinglian Jin and Yingju Jin, for their
unconditional love, trust, and support. Dad, Mom, brother and sisters, I love you all forever. I
would like to thank my brother and sisters for taking care of Mom during these years.
My deepest indebtedness also goes to Dad-in-law, Tulong Li and Mom-in-law, Lishui
Liu. They have not asked for anything from me but have given too much to me. I would also
thank my pretty sister-in-law, Zhizhen Li, who has helped me on many personal things at home.
Last but not least, my deepest love, gratitude, and indebtedness are for my wife, Zhiqing
Li, and my daughter, Joy L. Jin. Both of them are my sweethearts. Honey, I know both of you
are not living an easy life in a place so far away from me. I am sorry that I am not a good
husband and daddy. I love you.
viii
TABLE OF CONTENTS
APPENDIX A……..……………………………………………………………………….63
APPENDIX B……………………………………………………………………………...79
ABSTRACT……………………………………………………………………….………..xi
CHAPTER I: INTRODUCTION….......................................................................................1
CHAPTER II …………………………………………………………………………….…9
INTRODUCTION……………………………………………………………….…9
CHINA’S INTERGOVERNMENTAL FISCAL SYSTEM…………………….…14
LITERATURE REVIEW……………………………………………………….…19
EMPIRICAL STRATEGY………………………………………………………...23
EMPIRICAL RESULTS……………………………………………………..……29
CONCLUSIONS…………………………………………………………………..37
CHAPTER III…………………………………………………….…………………….…38
INTRODUCTION……………………………………………………………….. 38
INDIA’S INTERGOVERNMENTAL FISCAL SYSTEM…...…………………..42
LITERATURE REVIEW…………………………………………………………47
EMPIRICAL STRATEGY………………………………………………………..48
EMPIRICAL RESULTS………………………………………………………..…54
CONCLUSIONS…………………………………………………………………..57
CHAPTER IV: CONCLUSIONS………………………………………………………….58
REFERENCES………………………………………………………………………….…86
VITA……………………………………………………………………………………….96
ix
APPENDIX A
Table 1: China’s Statutory Revenue Assignments in 1993 and 2007…………………… 63
Table 2: China’s Statutory Expenditure Assignments in 1993 and 2007…………………64
Table 3: Variable Definitions for the China Regressions……...……………………….....65
Table 4: Summary of Previous Literature for China….…………………………………..66
Table 5: Summary Statistics for the China Sample (1985–2006)……………………...…67
Table 6: Regression Results for China (1985–2006 provincial data)…………………….68
Table 7: Specification Tests for the China Regressions…………………………….….…69
Table 8: Extreme Bounds Analysis for China…………………………………….………70
Table 9: Comparisons of Results for China………………………………………………71
Table 10: India’s Statutory Expenditure Assignments……………………………………72
Table 11: India’s Statutory Revenue Assignments……………………………………….73
Table 12: Variable Definitions for the India Regressions…………………………………74
Table 13: Summary Statistics for the India Sample (1980–2005)………………………..75
Table 14: Regression Results for India (1980–2005 state-wide data)……………………76
Table 15: Specifications Tests for India……………………………………………….…77
Table 16: Extreme Bounds Analysis for India……………………………………………78
x
APPENDIX B
Figure 1: Trend in the coefficients of variation for provincial expenditures per Capita in China, 1985–2006………………………………………………………………….....……….79
Figure 2: Shares of total revenues and expenditures in GDP and the share of central revenues in
total revenues, for China (1985–2006)……….……………………………………..…80 Figure 3: The non-linear effect of fiscal decentralization on growth in China…………………..81 Figure 4: Trends in the coefficients of variation for state expenditures per capita in India,
1980–2005………..…..………………………………………………….………….…82 Figure 5: Revenue and expenditure shares for the Union and the states of India,
1980–2005…………………………………………..………………………………....83 Figure 6: The non-linear effect of fiscal decentralization on growth in India…………………..84 Figure 7: The non-linear effect of fiscal decentralization on growth for China and India…..….85
xi
ABSTRACT
A COMPARATIVE STUDY OF FISCAL DECENTRALIZATION IN CHINA AND INDIA
BY
YINGHUA JIN
DECEMBER 2009
Committee Chair: Dr. Mark W. Rider
Major Department: Economics
This dissertation provides an empirical test of the effects of fiscal decentralization and
horizontal fiscal equalization on economic growth and examines the potential trade-off between
horizontal fiscal equalization and economic growth in both China and India.
Chapter II examines the effects of both fiscal decentralization and horizontal fiscal
equalization on economic growth in China, particularly the effect of the Tax Sharing System
reform enacted in 1994. Compared with previous studies, using more complete data providing
better measures and more econometrically sophisticated instrumental variable procedures, we
find that there is no substantial evidence of a trade-off between horizontal fiscal equalization and
growth. The 1994 Tax Sharing System reform has positively contributed to both economic
growth and horizontal fiscal equalization. In addition, we find that fiscal decentralization (FD)
has a non-linear effect on growth. For values of FD less than 21, fiscal decentralization has a
negative effect on growth, but for values greater than 21, fiscal decentralization has a positive
effect on growth.
Chapter III examines the effects of both fiscal decentralization and horizontal fiscal
equalization on economic growth in India, particularly the effects of the 1991 economic reforms.
Using state-wide data covering the period from 1980 through 2005, we find that fiscal
xii
decentralization has a negative effect on economic growth initially but that, beyond a certain
value of fiscal decentralization, the effect on growth becomes positive. However, further
decentralization could have a negative effect on horizontal fiscal equalization. These results are
robust. In the meanwhile, there is no evidence of a trade-off between horizontal fiscal
equalization and economic growth; instead, there is evidence of a positive effect of economic
growth on equalization. In addition, the 1991 economic reform has contributed to economic
growth.
A comparative study of China and India has shown that the degree of fiscal
decentralization in both countries is far from the point where its effect on economic growth
becomes positive. Despite the dangers of widening disparities in terms of interregional fiscal
resource distribution from further decentralization, no substantial evidence shows a trade-off
between horizontal fiscal equalization and growth in either country. An in-depth and more
thorough going fiscal decentralization with greater emphasis on equalization of fiscal disparities
are required in order to effect sustainable economic growth as well as social harmony in these
two Asian countries.
1
Chapter I: Introduction
As newly emerging economic powers on the global stage, China and India are the
subjects of great interest, especially in light of their remarkable record of economic growth in
recent years. As part of their adoption of market reforms, both countries are pursuing fiscal
decentralization reforms.1 Many students of public finance believe that fiscal decentralization
leads to greater allocative efficiency of the public sector and thus promotes economic growth. As
Robin Boadway (2006) points out, fiscal decentralization also may give rise to fiscal disparities
among sub-national governments due to interregional differences in local preferences, the cost of
providing services, revenue-raising capacities, and local fiscal policies among sub-national
governments. However, interregional fiscal disparities can be moderated by an
intergovernmental transfer system designed to equalize public-service provisions nationwide. In
this dissertation, we examine the effect of fiscal decentralization and horizontal fiscal
equalization on economic growth in China and India.
There are a variety of definitions of fiscal decentralization. According to Jennie Litvack
and Jessica Seddon (1999), fiscal decentralization is the transfer of fiscal authority and
responsibility from the central government to sub-national governments. Richard M. Bird (1993)
describes it in terms of top-down and bottom-up approaches. A top-down approach is the central
government delegating or decentralizing authority to local governments to increase the allocative
efficiency of the public sector. A bottom-up approach stresses the greater responsiveness of
government and greater opportunity for political participation in local government. In short,
1 By fiscal decentralization we mean the devolution of specific government functions by the central government to sub-national governments with the administrative authority and fiscal revenue to perform those functions.
2
fiscal decentralization does not have a precise definition and can be measured only with
difficulty and imprecisely.
There is a large theoretical literature describing the advantages of fiscal decentralization,
including increased allocative efficiency, increased productive efficiency, greater innovation, and
greater accountability of government to local residents. For example, Friedrich A. Hayek (1945)
contends that the aggregation of individual preferences for public goods by the central
government involves comparably greater costs than would occur with decentralization of
decision-making to local governments. In his seminal article on local government, Charles M.
Tiebout (1956) shows that ‘voting with your feet’ results in efficient allocation of fiscal
resources. George J. Stigler (1957) contends that representative democracy works best when
government is closest to the people. Wallace E. Oates (1972) contends that, in the absence of
scale economies and inter-jurisdictional spillovers, decentralized provision is always superior in
terms of economic efficiency to centralized and therefore uniform provision of local public
goods. Yingyi Qian and Barry R. Weingast (1997) describe the benefits for private property
rights of “market-preserving federalism”. Wallace E. Oates (1999) describes “laboratory
federalism,” in which best-practices can be identified by competing sub-national governments
trying novel economic experiments in a decentralized system of governance. Finally, John J.
Wallis and Wallace E. Oates (1988) and Timothy Besley and Stephen Coate (2003) contend that
sub-national governments can better tailor the provision of local public goods to diverse
preferences through decentralization, which is the preference-matching argument.
Although the literature identifies many potential benefits from fiscal decentralization,
there are some risks described in the literature, as well. Remy Prud'homme (1995) describes the
possible risks of fiscal decentralization, such as fragile preconditions for the potential gains,
3
regional disparities, and macro-economic instability. Govinda Rao and Nirvikar Singh (1999)
describe a “race to the bottom” in which competing sub-national governments engage in tax
competition to attract mobile factors of production, particularly capital, under provide local
public goods, and undermine the advantages of a domestic common market. Vito Tanzi (1996)
further cautions that fiscal decentralization could result in economic distortions due to excessive
regulation by sub-national governments and local corruption.
Since fiscal decentralization involves potential benefits and costs, the overall impact of
fiscal decentralization on economic growth is uncertain. What is more, the empirical evidence is
mixed, as well. Using provincial level data for the period 1980 through 1993, Tao Zhang and
Heng-Fu Zou (1998) find that fiscal decentralization has a negative effect on China’s economic
growth. In contrast, Justin Yifu Lin and Zhiqiang Liu (2000); Kunrong Shen and Wenlin Fu
(2005); Ying Ding (2005); and Baoyun Qiao, Jorge Martinez-Vazquez and Yongsheng Xu (2008)
find that fiscal decentralization has a positive effect on China’s economic growth. These studies
are discussed in greater detail below. To the best of our knowledge no one has examined the
effect of fiscal decentralization on India’s growth.
Before discussing the potential relationship between horizontal fiscal equalization and
economic growth, it is important to understand the concept of net fiscal residual which is the
difference between the benefit from public goods and services and their cost in terms of tax
burden. According to James M. Buchanan (1950), differences in net fiscal residuals across
regions may induce interregional migration from regions with low net residuals to regions with
high residuals. As people move to regions with high net fiscal residuals, the cost of providing a
constant level of local public goods increases due to increased congestion of local public goods,
and, as people leave regions with the low net fiscal residuals, the cost of providing a constant
4
level of the local public good decreases as congestion of the local public goods decreases. Thus,
in the Buchanan model, differences in net fiscal residuals among the regions of a country induce
migration which, in turn, eliminates the interregional differences. Thus, in equilibrium, there is
no incentive for people to move from one region to another due to differences in net fiscal
residuals.
One drawback to the Buchanan and Tiebout models is the assumption of a highly mobile
population. This assumption does not apply to very many countries. In particular it does not
apply to traditional societies like those in China and India. The lack of mobility in these
countries is often attributed to the strong attachments of people to particular localities due to
customs, such as ancestor worship in the case of China, and local languages and ethnic
fractionalization in the case of India. In contrast to Buchanan’s model, elimination in net fiscal
residuals among the regions of a country with an immobile population can only be achieved by
purposeful central government policies to transfer resources to eliminate such differences.
Another important rationale for horizontal fiscal equalization is provided by Paul Bernd
Spahn (2007), especially for countries like China and India. He points out that achieving social
cohesion and a sense of a national identity is particularly challenging in these two large countries.
Political unrest and regional separatist tendencies are on-going concerns. Indeed, there is concern
that growing interregional disparities within both countries are weakening national cohesion and
solidarity. There is a need for centripetal policies, such as equalizing transfers to help address the
concerns of aggrieved ethno-linguistic minorities and other potential secessionists.
In this study, we focus on horizontal fiscal equalization as a result of purposeful
government policy. As pointed out by Michael Smart (2007), the principle objective of
horizontal fiscal equalization is to eliminate differences in net fiscal benefits across regions
5
instead of reducing differences in individual incomes across regions. Thus, horizontal fiscal
equalization differs from the conventional role of government which is to redistribute income in
order to satisfy concerns about interpersonal equity. Horizontal fiscal equalization does not
ensure equitable treatment of individuals in similar circumstances. Instead, it seeks to equalize
the fiscal capacities of regional governments in order to achieve a balanced distribution of
resources among the regions.
As with fiscal decentralization, the effect of horizontal fiscal equalization on economic
growth is ambiguous. Robert M. Solow (1956); Gregory Mankiw, David Romer, and David N.
Weil (1992); Robert J. Barro and Xavier Sala-i-Martin (1995); Xavier Sala-i-Martin (1996); and
others model regional convergence in per capita incomes as the result of diminishing marginal
returns. Similarly, suppose the rate of return to a marginal increase in expenditures on
infrastructure and social services is higher in low-income regions than in high-income regions
due to diminishing marginal returns. Then, a central government policy of transferring fiscal
resources from high- to low-income regions will lead to a higher aggregate national growth rate.
Similarly, if equalizing transfers provide a social safety-net for the poor and thereby reduces civil
unrest arising from interregional fiscal disparities, then horizontal fiscal equalization would have
a positive effect on the aggregate national growth rate.
If, however, the taxes used to finance the equalizing transfer system are distortionary
and/or the transfer system itself is distortionary, then there may be a trade-off between horizontal
fiscal equalization and growth. Alternatively, low-income recipient regions may not use fiscal
resources as efficiently as high-income donor regions, perhaps due to differences in the quality
of governance among low- and high-income regions. The result would be a trade-off between
equalization and growth.
6
This dissertation focuses on the effect of two major policies on growth in China and India,
specifically fiscal decentralization and horizontal fiscal equalization. There are important reasons
to focus on the effect of these two policies on economic growth in China and India. On the one
hand, both countries are experiencing high economic growth rates. On the other hand, the rural
poor in both countries are not fully sharing in the benefits of rapid economic growth. In part,
expenditure decentralization reforms are meant to empower local governments to pursue
economic development that best suits local preferences and comparative advantage. Horizontal
fiscal equalization is a means of addressing concerns about growing interregional disparities in
the quality of public services that may result from fiscal decentralization due to differences in
regional attributes and policies. Therefore, examining the combined impact of fiscal
decentralization and horizontal fiscal equalization on economic growth is an important policy
concern in both China and India.
This dissertation intends to enrich our limited knowledge of the effects of fiscal
decentralization and horizontal fiscal equalization on economic growth. There are important
reasons to focus on China and India. First, China and India are the two fastest-growing and
largest emerging economies in the world. Eswar Prasad and Thomas Rumbaugh (2004) calculate
that, between 2001 and 2003, China accounted for almost one-quarter of global growth in gross
domestic product (GDP) when output is measured using purchasing power parity exchange rates.
Second, both countries are pursuing a gradualist approach to economic reform as opposed to the
“big-bang” approach to reform adopted in many countries of the former Soviet Union, Eastern
Europe, and Indonesia. Third, both central governments have been devolving more and more
power, resources, and responsibility for service delivery to sub-national governments.
7
Despite these remarkable similarities, there are interesting contrasts between the two
countries, as well. China is the world’s largest unitary country in terms of population, while India
is the world’s largest federal country. India is ethnically diverse; while China’s population is
relatively homogeneous. China is governed by one-party rule, while India is the world’s most
populous democracy, with competitive multi-party elections. This comparative study should shed
light on the performance of their decentralization policies in terms of their effects on economic
growth in China and India. The size and importance of these two developing countries in and of
itself should make this comparative study interesting to students of public finance and to students
of economic develop. The results of this study may also provide useful lessons to other
developing countries considering or currently pursuing reforms of their fiscal system.
We find that fiscal decentralization has a non-linear effect on economic growth in both
countries. For low values of fiscal decentralization, the effect on growth is negative. Beyond a
value of approximately 20, fiscal decentralization begins to have a positive effect on growth. It is
somewhat surprising that the point where fiscal decentralization switches from having a negative
effect on growth to a positive effect is nearly identical in both countries: the value is between 20
and 21 for China and India, respectively. It is also interesting to note that the negative (positive)
effect of fiscal decentralization on economic growth is stronger for India compared to China, in
both absolute terms and relative to GDP per capita. In contrast to Baoyun Qiao et al. (2008), we
also find no substantial evidence of a trade-off between economic growth and horizontal fiscal
equalization for either country.
The remainder of this dissertation is organized as follows. Chapter 2 re-examines the
potential trade-off between horizontal fiscal equalization and growth in China, Chapter 3
8
presents an examination of the potential trade-off between horizontal fiscal equalization and
growth in India, and Chapter 4 concludes.
9
Chapter II: Re-Examining the Potential Trade-Off between
Horizontal Fiscal Equalization and Growth in China
Introduction
During the past quarter of a century, China has experienced rapid growth in per capita
income in both absolute and relative terms. In 1980, at the outset of China’s opening to global
trade and market reforms, China’s per capita income was $192 in purchasing power parity (PPP)
terms, which ranked 141st among approximately 145 countries.2 By 2005, China’s per capita
income was $1,720 (PPP), which ranked 108th in the world. This impressive record of economic
growth also reduced income inequality among China’s provinces. In 1981 Shanghai had the
highest per capita income of all provinces in China, with a gross regional product (GRP) of $808
per capita.3 The poorest sub-national jurisdiction was Guizhou, with a GRP of $44 per capita.
Thus, Shanghai’s per capita income was more than 18 times that of Guizhou’s at the outset of the
reforms.4 By 2005, however, Shanghai’s income per capita was 10 times that of Guizhou, or
$7,533 versus $776, respectively. Although there is clearly substantial interregional income
inequality in China, these figures reflect substantial progress in reducing interregional income
inequality during this period.
As Figure 1 shows, the coefficient of variation (CV) of per capita provincial expenditures
was about 79 percent in 1985 and decreased to 54 percent in 1993. In 1994, as discussed in
greater detail below, China adopted the TSS reforms to equalize public expenditures among the
provinces. During the following decade, as a result of TSS, the CV of per capita provincial
2 According to the CIA World Factbook and World Development Indicator Database, the dollar figures for GDP are converted from domestic currencies using annual exchange rates. 3 We use the RMB-U.S. dollar exchange rate, which was 0.146321 on April 12, 2009 (www.xe.com/ucc/convert.cgi). 4 China Statistical Yearbook (1982, 2005).
10
expenditures has been increasing and peaked at 82 percent in 2004. After 2004, the CV of per
capita expenditures sharply declined to 62 percent in 2006, as a result of a significant increase in
transfers during those years. As a whole, the equalization impact of the transfer system is rather
modest, as evidenced by the effect of the 1994 TSS reforms on the CV of per capita expenditures
pre- and post-1995.
Without central government transfers, however, the interregional disparities in per capita
expenditures would have been much greater. As Figure 1 shows, the CV of per capita
expenditures, without transfers and deficits, varies between 122 percent in 1997 to 540 percent in
2004. These figures are significantly greater than the peak CV of per capita expenditures after
transfers of 82 percent in 2004. Even after transfers, though, disparities in per capita
expenditures remain large. Figure 1 clearly shows that central government transfers and
provincial borrowing are having an equalizing effect on provincial expenditures. As previously
noted, the effect of provincial borrowing on interregional fiscal disparities is often ignored in the
literature on China due to the lack of data.
Bert Hofman and Susana Cordeiro Guerra (2005) also find that disparities in provincial
expenditures per capita are declining as a result of the transfer system. They note that the income
elasticity of provincial expenditures per capita is less than the income elasticity of provincial
revenues per capita. These finding imply that the transfer system is having an equalizing effect
on provincial per capita expenditures, which is consistent with the patterns observed in Figure 1
and summarized above. Nevertheless, the CV of provincial expenditures per capita, after the TSS
reforms, remains large. We show below that some components of the central transfer system are
equalizing, while others are counter-equalizing. Another point discussed in greater detail below
is that provincial government borrowings in China are having an equalizing effect on provincial
11
per capita expenditures. This is a fact often overlooked in the literature on China’s local
government finances.
The question that arises from the foregoing analysis is whether China’s twin policies of
fiscal decentralization (FD) and horizontal fiscal equalization (HFE) are growth promoting. From
a theoretical perspective, we contend that the effects of FD and HFE on economic growth are
ambiguous.
There are a number of reasons to expect a trade-off between HFE and growth. Generally
speaking, HFE involves shifting resources from high- to low-income regions. This transfer of
resources may reduce public infrastructure investments in high-income regions and presumably
increases pubic infrastructure investments in low-income regions. In low-income regions,
however, capital investments may not necessarily increase by the same magnitude as the
decrease in high-income regions.
First, according to Per-Olov Johansson (1991), the central government must use
distortionary taxes to finance the intergovernmental transfer system. The hypothesized positive
effect of equalization transfers on the national growth rate due to diminishing marginal returns
and the negative effect on growth of distortionary taxes would have an offsetting effect on
growth. Depending upon which effect is stronger, HFE could have a positive or negative effect
on the aggregate national growth rate.
Second, Charles E. McLure, Jr. (1995) points out that the intergovernmental transfer
system itself may be distortionary -- like Arthur Okun's (1975) leaky bucket – which may result
in a trade-off between HFE and growth. Third, if the transferred resources cannot be utilized by
low-income regions as efficiently as by high-income regions, then HFE may have a negative
effect on economic growth.
12
In contrast, a positive relationship between HFE and economic growth could arise. For
example, horizontal fiscal equalization may reduce social instability and create an investment
environment that is more favorable to economic growth. Paul Bernd Spahn (2007) contends that
social cohesion is very important in a country like China where political and ethnic unrest and
regional separatist’s movements are important concerns. Narrowing interregional fiscal
disparities through a purposeful policy of HFE could moderate social unrest and thereby create a
better climate for investment. At the same time, equalizing transfers could be used to finance
improved education and health in low-income provinces and thereby promote human capital
formation among the poor. In an endogenous growth model, Kevin James Bowman (2007)
shows that initial educational inequality reduces the subsequent economic growth rate. Therefore,
reducing initial educational inequality through a policy of HFE should have a beneficial effect on
economic growth.
In addition, according to the growth model of Robert M. Solow (1956), the rate of return
to public capital per capita may be higher in low-income provinces than in high-income
provinces due to diminishing marginal returns. In this case, a policy of transferring fiscal
resources from high- to low-income provinces may have twin virtues: equalizing per capita
expenditures and increasing the aggregate growth rate. If the transfers can be used more
efficiently by low-income regions, and if there are higher returns to capital in low-income
regions, then HFE should have a positive effect on the national growth rate, even in the presence
of Okun’s leaky bucket. In short, we contend that, from a theoretical perspective, the effect of
fiscal equalization on the rate of economic growth is ambiguous.
13
Finally, Georgia Guess, William Loehr and Jorge Martinez-Vazquez (1997); Elliott
Parker and Judith Thornton (2007); and Baoyun Qiao, Jorge Martinez-Vazquez and Yongsheng
Xu (2008) find empirical evidence of a trade-off between HFE and growth.
The purpose of this study is to examine the effect of China’s policy of FD and HFE on
growth. Following the work of Qiao et al. (2008), we estimate a simultaneous equations model
(SEM). In the growth equation, we control for many of the conventional growth determinants—
such as capital investment, labor growth, openness, agriculture share, inflation rate, foreign
direct investment, natural catastrophes, and the 1994 TSS reforms. Furthermore, we account for
the potential endogeneity of FD, HFE, and growth, using an instrumental variables approach. We
estimate a variety of models, using provincial data for the period 1985 to 2006.
This paper improves upon the existing literature in the following ways. First, we use
annual provincial data for the period 1985 to 2006, which straddles the enactment of the
important 1994 TSS fiscal reforms. More specifically, these data include nearly all time periods
examined in previous studies and more completely covers the pre- and post-1994 TSS reform
period. Second, we employ better measures of FD and HFE. In contrast to Qiao et al. (2008), our
measures account for extra-budgetary funds not only of sub-national governments but also those
of the central government. In contrast to others, our HFE measure additionally accounts for the
equalization policy pursued by China’s central government after 1994. Third, we account for the
potential endogeneity of FD, HFE, and growth over time and across provinces. Finally, we
exploit the panel nature of our data by estimating fixed effects models (FE), two-stage least
squares models with fixed effects (2SLS FE), and two-step generalized method of moments
models (GMM).
14
We find that FD has a non-linear effect on growth. For values of the fiscal
decentralization index ranging from 0 to 20, the effect of FD on growth is negative. For values of
the index greater than 20, the effect is positive. This finding is robust to alternative econometric
specifications and time periods. In contrast to Baoyun Qiao et al. (2008), we find no substantial
evidence in favor of a trade-off between HFE and growth. Finally, we find that the 1994 TSS
reforms are having a positive effect on economic growth. Meanwhile HFE improved modestly
after the enactment of the 1994 TSS reforms. In part, though, the increase in HFE is the result of
increased provincial borrowings rather than the implementation of an equalizing transfer system.
The remainder of this chapter is organized as follows. The next section provides some
background on China’s intergovernmental fiscal system. In the third section, we summarize the
previous literature on the effects of FD on China’s growth. The fourth section describes the
empirical strategy, data, and variable construction. In the fifth section, we discuss the empirical
results, and the final section concludes.
China’s Intergovernmental Fiscal System
This section describes the evolution of China’s fiscal decentralization reforms; elaborates
on China’s intergovernmental fiscal system, including China’s four pillars of fiscal
decentralization, with special emphasis on the 1994 TSS reforms and the transfer system
introduced in 1994; and briefly describes China’s system of extra-budgetary finance.5
Institutionally and legally, the People’s Republic of China has a unitary form of
government with five levels of government hierarchically arranged in a pyramid-like fashion
5 The four pillars of fiscal decentralization are the following: expenditure assignments, revenue assignments, intergovernmental transfer system, and regulation of sub-national borrowing autonomy. For a good discussion of the four pillars of decentralization in the context of the United Republic of Tanzania, see Jameson Boex and Jorge Martinez-Vazquez (2006).
15
with the central governmental naturally at the peak.6 The recent evolution in the direction of
greater FD has been a gradual process. In 1978, fiscal reforms started with the devolution of
control over resources and decision-making power to sub-national governments as well as to
state owned enterprises (SOEs).
Although the fiscal reform initiatives have provided sub-national governments with
incentives to develop their local economy, the uncontrolled decentralization and case-by-case
bargaining between the central and sub-national governments introduced by several fiscal reform
packages in the 1980s and early 1990s led to a sharp decline in total central government revenues.
As a result of arbitrary tax exemptions granted to local SOEs and the shifting of taxes from
budgetary revenues to extra-budgetary revenues by sub-national governments, as shown in
Figure 2, the share of central government revenues in GDP dropped from 22 percent in 1985 to
12 percent in 1993, and the central government’s share of total revenues dropped from 38
percent in 1985 to 22 percent in 1993.
These two declines in revenue shares caused the central government to initiate the 1994
TSS reforms and to begin re-centralizing revenue collections in 1995. According to Zhihua
Zhang and Jorge Martinez-Vazquez (2003a), the purpose of the 1994 TSS reforms was to
increase total fiscal revenue as a share of GDP and to increase the share of central government
revenue in total fiscal revenues. As summarized in Table 1, several important sources of tax
revenues, like the value-added tax (VAT), the corporate income tax (CIT), and the personal
income tax (PIT), were shared taxes before 1995, with the exception of the CIT which only
became a shared tax after 1994.
6 See Figure 4.1 of Baoyun Qiao and Anwar Shah (2006) for the structure of government in China and its political system.
16
After the 1994 TSS reforms, the central government takes the major share of these taxes.
Taxes exclusively assigned to sub-national governments are low-yielding types such as slaughter
taxes. Meanwhile, tax administration was divided into two systems: the National Tax Service
(NTS) and the Local Tax Service (LTS). Wanda Tseng, Hoe Ee Khor, Kalpana Kochhar,
Dubravko Mihaljek and David Burton (1994) report that NTS is responsible for the collection of
exclusively central and shared taxes. The LTS exclusively collects local revenues. As a result of
this clear-cut division, fiscal decentralization prevents sub-national governments from hiding
local resources in black boxes.7
As a result of the 1994 TSS reforms, the four pillars of China’s intergovernmental fiscal
system are characterized by highly decentralized expenditures, highly centralized revenues, a
high degree of sub-national government transfer dependence, and hidden sub-national
borrowings. The previously poor inland provinces became even worse off in terms of revenue
capacity because their limited own revenues were centralized. This makes them even more
dependent on central government transfers as well as on hidden sub-national borrowings to
finance their operations.
A fiscal system with highly centralized revenues and highly decentralized expenditures
makes reliance on intergovernmental transfers unavoidable. The 1994 TSS reforms introduced
the current transfer system. This system consists of three types of intergovernmental transfers:
tax rebates, general-purpose grants, and special-purpose grants. To address the political
opposition of high-income sub-national governments to the TSS reforms, the system of tax
7 Before the TSS reform in 1994, sub-national governments had an incentive to provide tax exemptions to local SOEs on an ad hoc basis and to shield the exempted tax revenues from the central government; however, after the separation of the two tax administrations by TSS, sub-national governments neither have incentives nor ability to make them. For details about sub-national government’s use of tax exemptions, see Wanda Tseng (1994).
17
rebates are derivation based. Accordingly, tax rebates conflict with the equalization goal of the
1994 TSS reforms. To achieve greater horizontal fiscal equalization, the share of tax rebates was
decreased from 75 percent in 1994 to 19 percent in 2008, while that for general purpose grants
increased from 4 to 38 percent during the same period. Special-purpose grants compromise a
third type of intergovernmental transfer and are intended to realize specific objectives of the
central government. The share of special-purpose grants in total transfers also increased from 21
percent in 1994 to 43 percent in 2008. In short, the equalization purpose of the transfer system is
compromised by the counter-equalizing effect of tax rebates.
Although sub-national governments are prohibited from borrowing except upon special
approval of the central government, almost all sub-national governments are circumventing these
restrictions by borrowing “off the books.” Baoyun Qiao and Anwar Shah (2006) describe the
“off the books” channels of local government borrowings. Examples of such channels include
borrowings from local commercial banks by sub-national government-controlled enterprises
(such as local SOEs) with local government guarantees; establishment of new companies through
collective financing by local governments or employee’s of local SOEs; borrowing from pension
funds, unemployment insurance funds, or government employee salary funds; borrowing from
extra-budgetary funds, and so on. Interestingly, these sub-national borrowings are having a
partial equalizing effect on provincial expenditures per capita, as shown by the square-dotted line
and the triangle-dotted line in Figure 1.
Although there are extensive concurrent expenditure responsibilities, more than 74
percent of national revenues were spent by sub-national governments in 2005 as opposed to 47.5
percent in 1984. As shown in Table 2, many expenditure responsibilities are shared by central
and sub-national governments, including social security, education, medical services, public
18
health, and economic development. A heavy burden of providing basic public services rests
largely on sub-national governments. More specifically, 98 percent of health expenditures were
shouldered by sub-national governments in 2007. Furthermore, the politically centralized system
makes it easy for the central government to shift expenditures downward to sub-national
governments, although major revenues have been recentralized since enactment of the 1994 TSS
reforms.
Table 1 shows tax and non-tax revenue shares in total revenues for central and sub-
national governments before and after the TSS reform. The revenue recentralization evident in
Table 1 is mainly the result of the 1994 TSS reforms. As a result, the percentage of shared tax
revenues in total tax revenues increased from 55 percent in 1994 to 70 percent in 2005. Before
the 1994 TSS reforms, the share of sub-national tax revenues in total revenues was 79 percent in
1993, with the remaining 21 percent of revenue accruing to the central government. However, in
2007, the situation was reversed. In 2005, central government transfers to sub-national
governments accounted for 57 percent of total central expenditures and 46 percent of total sub-
national expenditures after accounting for extra-budget accounts.
For the case of China, there exists another critical source of finance. That is off-budget
finance, which consists mainly of extra-budgetary funds together with user fees and various local
levies. Off-budget finance constitutes a fiscal dual track, together with budget finance and sub-
national extra-budgetary accounts and is under the full discretion of sub-national governments.
The size of extra-budgetary funds grew rapidly in the 1980s and 1990s, finally becoming
equivalent in size to budgetary funds in 1991.8 As a result of several policy reforms that shifted
8 See Christine Wong (1998) for the size evolution and detailed categories of off-budget funds.
19
extra-budgetary funds to budgetary funds, extra-budgetary funds started to decline.9 In 2005,
extra-budgetary expenditures dropped to about 16 percent of budgetary expenditures on average.
Since the off-budget funds are levied on the same tax base as budget funds, off-budget funds,
like extra-budgetary revenues, are positively related to own source revenues, which is an
important source of interregional disparities in per capita provincial expenditures.10
Literature Review
This section begins by discussing some issues with measuring fiscal decentralization
which is an obvious challenge in studies like this one; then we proceed with a discussion of the
evidence of HFE in China. Finally we conclude this section with a summary of previous findings
on the effect of FD and HFE on growth in China.
Due to the lack of a precise definition of fiscal decentralization, there are many studies
offering operational definitions of FD. Nevertheless, there is no conclusive result: no measure is
obviously superior to all others. The criticism is that no single scalar measure of FD can capture
the entire panoply of sub-national fiscal discretion and control. In any event, two measures are
frequently used to measure the degree of fiscal decentralization in a country: an expenditure
measure and a revenue measure. The expenditure measure is the ratio of sub-national
government expenditures to consolidated government expenditures. A larger value of this ratio
indicates a greater degree of expenditure decentralization and thus greater fiscal decentralization.
The revenue measure is the ratio of sub-national own-source revenues to consolidated
government revenues. Again, a greater value of this ratio indicates greater revenue
decentralization and thus greater fiscal decentralization.
9 See, for example, State Council Document No. 29 (1996). 10 See, for example, Christine Wong (1998) for detailed empirical investigation of extra-budgetary funds and Xuejin Zuo (1996) for local levies, among others.
20
Many students of public finance believe that the ‘true’ measure of fiscal decentralization
is best captured by revenue decentralization. For example, Robert D. Ebel and Serdar Yilmaz
(2002) contend that for there to be genuine fiscal decentralization, local governments must have
the authority to tax their “own sources.” Roy W. Bahl (2005) also stresses that the “essence” of
decentralization is the empowerment of sub-national governments to raise their own revenues by
having control over the tax rate of at least one broad-based tax (i.e., property tax, income tax,
consumption tax).
In the case of China, none of these requirements of revenue decentralization are
completely satisfied. Sub-national governments are not allowed to introduce new taxes or change
tax rates on major tax bases, although they can make full use of their tax capacity through efforts
to grow their tax bases or strengthen tax collections.11 In the case of China, the degree of revenue
decentralization is very limited, so the literature tends to focus on the degree of expenditure
decentralization. Shaoguang Wang (1997) contends that China is over-decentralized according to
the expenditure measure of decentralization because more than 70 percent of national
government spending is conducted by sub-national governments.
We follow the conventional approach by using expenditure decentralization to gauge the
degree of fiscal decentralization in China. We believe that this measure best captures the
evolution of fiscal decentralization in China and has the added benefit of allowing us to compare
11 Roy W. Bahl (1999) summarizes several ways in which sub-national governments in China have acted against centralized taxation power, which he calls “backdoor federalism”. Provincial governments shift tax revenues from budget accounts to extra-budgetary accounts; arbitrarily exempt taxes of sub-national SOEs and then cooperate with them under the table to finance sub-national spending; influence banks at the sub-national level to grant policy loans to many unqualified projects; infiltrate the national tax administration at the sub-national level to change the division of tax revenues between the central and sub-national governments; levy informal surcharges and fees on public goods and services; and so on.
21
our results with previous findings in the literature.12 As previously noted, our measure of
expenditure decentralization is more comprehensive than those used in the existing literature
because our data allow us to account for both central and provincial government extra-budgetary
expenditures.
Turning to HFE in China, there is substantial evidence of the equalizing effect of China’s
transfer system. For his analysis of the effects of equalization grants in China, Bo Zhang (2006)
uses an absolute indicator, which is the gap between Gini coefficients for per capita revenues and
per capita expenditures; a relative indicator, which is the ratio of the absolute indicator to the
Gini coefficient of per capita fiscal revenue; and the Theil index of per capita provincial revenue
and expenditure. He finds that fiscal transfers have only partially achieved their equalization goal,
although inequality in per capita provincial expenditures is less pronounced than before the 1994
TSS reforms. David Dollar and Bert Hofman (2008) find that the transfer system has a limited
equalization impact using budget data -- they do not account for extra-budget data -- in fiscal
year 2003. Lou Jiwei and Ehtisham Ahmad (1997); Jiwei Lou (2008); Christine Wong (2009);
and Zhihua Zhang and Jorge Martinez-Vazquez (2003b) all conclude that the reforms have not
entirely achieved the intended goal of equalizing provincial per capita expenditures.
Re-examining the effects of fiscal decentralization and fiscal equalization on economic
growth is quite important. As Jorge Martinez-Vazquez and Robert McNab (1997) point out in
their review of the literature on fiscal decentralization, the state of knowledge about the effects of
fiscal decentralization is extremely limited. Most existing empirical studies examine the
relationship between fiscal decentralization and economic growth; only a few have examined the
effect of fiscal equalization on growth. Furthermore, this literature provides conflicting evidence
12 See Kunrong Shen and Wenlin Fu (2005), Andrew Feltenstein and Shigeru Iwata (2005), Nobuo Akai and Masayo Sakata (2002), Tao Zhang and Hengfu Zou (1998), etc.
22
on decentralization’s effect on growth. Finally, there is limited empirical evidence on the
potential trade-off between fiscal equalization and growth. The exception is Qiao et al. (2008)
who use a panel of provincial data for the period 1985 through 1998. They find evidence of a
trade-off between horizontal fiscal equalization and economic growth. One limitation of their
study is that their data only include the initial years of the 1994 TSS reforms.
Table 3 provides a list of variable definitions, and Table 4 summarizes previous empirical
results on the impact of fiscal decentralization and fiscal equalization on economic growth in
China. Turning to Table 4, Tao Zhang and Heng-Fu Zou (1998) find that fiscal decentralization
has a negative and statistically significant effect on economic growth. However, none of the
conventional variables in growth equations, like capital and labor, is statistically significant at
conventional levels.
Justin Yifu Lin and Zhiqiang Liu (2000) and Ying Ding (2005) find that fiscal
decentralization has a positive effect on growth, using data for the periods 1970-1993 and 1994-
2002, respectively. However, their empirical strategies do not account for the potential
endogeneity of fiscal decentralization in the growth equation. Finally, Baoyun Qiao et al. (2008),
using data for the period 1985-1998, find that fiscal decentralization has a non-linear effect on
growth. For small values of the fiscal decentralization index (0 < FD < 0.5), the effect on growth
is positive but the effect becomes negative for values of FD greater than 0.5. The mean value of
FD in their sample is 0.7, which implies that on average fiscal decentralization is having a
negative effect on economic growth in China, at least during the period spanned by their data.
The data used in these studies do not provide adequate coverage of the periods both
before and after enactment of 1994 TSS reforms. These reforms alone are sufficiently important
to merit further study. They also provide exogenous variation in fiscal decentralization and
23
horizontal fiscal equalization that may be useful in identifying their effects on China’s economic
performance. We proceed below with a description of our empirical strategy.
Empirical Strategy
To examine the effect of fiscal decentralization and equalization on growth, we estimate
two equations: a growth equation and an equalization equation. For the growth equation, we
adopt a production-function-based model that has been widely used in the empirical literature on
growth, including the studies of Robert M. Solow (1956), Trevor Swan (1956), Gregory Mankiw,
David Romer and David N. Weil (1992), Hamid Davoodi and Heng-Fu Zou (1998), among
others. For the equalization equation, we include variables that economic intuition suggests may
influence equalization.
Following Baoyun Qiao et al. (2008), we estimate the following system of equations:
itiitit
ititititititit
ititititititititit
ASIACTI�
CATDTSSGOVSCDEVTETEAGS
FDIRFLKFDFDEG
υµαα
ααααααα
ααααααααα
++++
+++++++
++++++++=
1716
15141312
2
11109
87654
2
3210
(1)
itiititititit
itititititit
ititititititit
ASIACTI�CATDTSSXBGT
L�POPMWSGOVSCDEVTETE
AGSFKFDFDGE
εδααβββ
ββββββ
βββββββ
+++++++
++++++
++++++=
1716151413
1211109
2
87
654
2
3210
(2)
The subscript i is the observation unit, province; t is the time unit, year. G is the growth
rate in real GRP per capita. E is an index of equalization and follows the definition of Qiao et al.
(2008). More specifically, we compute the difference in year t between a province’s per capita
expenditures in year t and the mean of provincial expenditures in year t, for every province and
year in our sample. We convert these differences into ‘distances from the mean per capita
provincial expenditures in year t’, by taking the absolute value of these figures. Then, we
24
normalize these year t ‘distances’ by dividing each of them by the year t mean per capita
provincial expenditures. Finally, we apply a minus sign to these figures in order to give our index
of equalization (E) a more intuitive interpretation. As E approaches zero from the left, the
province’s per capita expenditures are becoming closer to the mean of per capita provincial
expenditures and thus becoming more equal.
Following Tao Zhang and Heng-Fu Zou (1998), the degree of fiscal decentralization (FD)
is the ratio of provincial government per capita expenditures and central government per capita
expenditures. For this purpose provincial (central) government expenditures include both
provincial (central) budgetary and extra-budgetary expenditures. Including FD and FD-squared
in the model allows the degree of fiscal decentralization to have a non-linear effect on the
dependent variable.
The growth equations include a vector of control variables that are typically used in
empirical growth models. Capital (K) is measured by the growth rate of investment in total fixed
assets. Labor (L) is measured by the growth rate in employment. According to economic theory,
K and L should have a positive effect on growth. Openness to international trade (F) is measured
by the share of exports plus imports in GRP. Previous studies, like those by Xiaojuan Jiang
(2004) and Kunrong Shen and Jian Li (2003), find that the degree of openness promotes
economic growth. R is the percentage change in the general retail price index. On the one hand,
Tao Zhang and Heng-Fu Zou (1998) contend that a higher inflation rate can encourage people to
invest more in capital and reduce real balance holdings which is the Tobin portfolio-shift
effect.13 In this case, inflation would promote economic growth. On the other hand, a higher
13
See James Tobin (1956) for a detailed discussion.
25
inflation rate increases the costs of consumption and investment and thus may impede economic
growth.
FDI is the natural logarithm of per capita foreign direct investment, which also is
expected to have a positive effect on growth. In addition, we include the share of agriculture in
total GRP (AGRS), which is expected to have a negative effect on economic growth. Provincial
tax effort (TE) is the share of provincial revenues in GRP. Provincial revenues include the
revenue from all shared taxes that go to sub-national governments, including tax rebates, as well
as local taxes, transfers, and non-tax revenues. TE is expected to have a negative effect on
growth since taxes distort the economic behavior of consumers and firms. We include the square
of TE because of the potential non-linear effect of tax effort on the dependent variable.
Infrastructure investment (CDEV) is equal to central government development expenditures.
Provincial government size (GOVS) is the share of provincial budgetary and extra-budgetary
expenditures in GRP. Economic theory and previous empirical work suggest that these two
variables should have a positive effect on economic growth.
TSS is a vector of time dummy variables set equal to “1” after 1994 when the TSS
reforms became effective. Given the goals of these reforms described above, we expect this
variable to have a positive effect on growth. CATD is a set of natural catastrophe dummy
variables that are expected to have an adverse affect on agricultural production. CATD is set
equal to “1” for the province and in the year in which the catastrophe occurs and “0” otherwise.
TIN is a time dummy variable set equal to “1” in 1989 which is the year of the Tian’anmen
Square incident and “0” otherwise; ASIAC is a time dummy variable set equal to “1” in 1997 --
the year of the Asian financial crisis -- and “0” otherwise. Both variables are included to account
26
for the adverse economic shocks of these two events and are expected to have a negative effect
on growth.
In the equalization equation, we include the growth rate (G), fiscal decentralization (FD),
and FD-squared. As previously discussed, the effect of growth on equalization is ambiguous.
Fiscal decentralization is likely to give rise to interregional disparities. Therefore, FD is likely to
have a negative effect on equalization. We include FD-squared to allow fiscal decentralization to
have a potentially non-linear effect on equalization. Capital (K), openness (F), and share of
agriculture (AGS) are expected to increase fiscal disparities. TE captures the effect of tax effort
on equalization. As discussed by Tao Zhang and Heng-Fu Zou (1998), tax effort is expected to
have a positive effect on equalization (E) because greater tax effort, everything else held constant,
should result in greater spending per capita. CDEV are equalizing transfers and therefore should
have a positive effect on equalization. The formula used to distribute development funds ensures
that this variable is exogenous. Government size (GOVS) is expected to have a negative effect
on equalization because a larger government crowds-out private investment which impedes
economic growth, narrows the tax base, and lowers tax revenues.
We include the share of missing women in the total population (MWS). Using the
method of Ansley J. Coale and Judith Banister (1994), we calculate the number of hypothesized
missing women for each province and in each year in our sample. MWS is expected to have an
adverse effect on equalization. For example, increased pensions are required for elderly men who
never marry because of China’s adverse sex ratio that results from the cultural preference in
China for a male child which many believe is exacerbated by China’s one child only policy.14
LNPOP is the natural logarithm of the provincial population which is expected to have a positive
14 See, for example, Quanbao Jiang, Marcus W. Feldman and Xiaoyi Jin (2005) for the adverse effects of missing women.
27
effect on equalization. There is a widespread belief among students of China’s local public
finances that extra-budget funds are a main contributor to interregional fiscal disparities in
China.15 Thus XBGT, defined as the ratio of extra-budget expenditures and budgetary
expenditures, is included to capture this effect. Given the goals of the TSS reform, we expect this
variable to have a positive effect on equalization. CATD, TIN, and ASIAC are expected to
increase fiscal disparities.
The terms µi and δi in equations (1) and (2), respectively, are unobserved, time-invariant
provincial effects on the dependent variable. The terms υit and εit, in (1) and (2) respectively, are
idiosyncratic shocks that are time-varying and represent unobserved factors that change over
time and affect the dependent variable. In this SEM, a change in any disturbance term of
equation (1) leads to a change in the potentially endogenous variable E that it directly determines.
This, in turn, changes the other potential endogenous variable, G. Similar logic applies to the
effects of a change in any disturbance term in equation (2).
The data used in this study come from the China Data Center of the University of
Michigan and covers the period from 1985 through 2006. The data on extra-budgetary revenues,
extra-budgetary expenditures, and central transfers come from the Ministry of Finance (retrieved
on October 2008 from www.mof.gov.cn). This period covers the main years of previous studies
and straddles the 1994 TSS reform. The extra-budgetary data also are available for this period.
Naturally, there is concern that China’s government may be manipulating economic data
for political reasons. According to Thomas G. Rawski (2001), the intentional falsification of data,
in terms of GRP statistics in particular, is common at all levels of government in China. In
contrast, Gregory Chow (2006) concludes that China’s official statistics are generally reliable
15 See, for example, World Bank (2000) and World Bank (2001).
28
and consistent with China’s economy, although some data must be used with caution, as with
any other data for any other country. In particular, the extra-budgetary data from the Ministry of
Finance is considered to be reliable.
Our sample spans the years from 1985 to 2006 and includes data on 31 provinces of
mainland China, including Tibet, but does not include data for the island governments of Hong
Kong, Macau, or Taiwan. In 1997, the municipality of Chongqing separated from Sichuan
Province, and in 1988 Hainan separated from Guangdong Province. Data for Chongqing and
Hainan Provinces are available for the years after the bifurcations created these new provinces in
1997 and 1988, respectively. In estimating the two-step GMM, we lose a year of data.
Consequently, there should be 638 observations in our data: 29 provinces over a 21 year period,
Hainan Province for 19 years, and Chongqing Province for 10 years (29 × 22 = 638). However,
we must drop 8 observations due to missing values, resulting in a sample consisting of 630
observations. In the case of the other studies, they lose many more observations as a result of
missing values. So, again, we believe that our data have advantages over those used in previous
studies.
Table 5 provides summary statistics for our sample of 630 observations. The average
growth rate in real GRP per capita (G) in our sample is 9.45 percent per annum, the maximum
growth rate is 38.1 percent, and the minimum is -3.7 percent. The CV of the provincial growth
rates in our sample is 43.5, meaning that the standard deviation of the growth rates is 43.5
percent of the mean growth rate in our sample. The degree of fiscal decentralization (FD) ranges
from 0.86 to 17.85, with a mean value of 3.34. Horizontal fiscal equalization (E) varies between
-3.46 and 0.0, with a mean value of -0.46. The average growth rate of fixed asset investment (K)
is 20.46 percent per annum, which is substantial, and the average growth rate of employment (L)
29
is 1.84 percent per annum. The average rate of inflation (R) is 5.86 percent, and the average
share of agriculture in GRP (AGS) is 21.4 percent. The average share of provincial revenues in
GRP (TE) is 14 percent, and the average share of extra-budgetary funds in budget funds (XBGT)
is 0.44 percent. Finally, the average share of missing women is 4 percent of the provincial
population, and varies between 1 and 6 percent in our sample. Now, we turn to a discussion of
our empirical results.
Empirical Results
This section begins by reporting the estimation results for a variety of specifications.
Second, we carry out detailed specification tests. Third, we conduct an “extreme bounds
analysis” (EBA) for the growth equation to justify our selection of the baseline model reported in
Table 6. Fourth, we use data for the period 1949 through 2008 and replicate the specification of
Qiao et al. (2008) to further check the robustness of our results.
Table 6 provides the main results. We begin by discussing fixed effects (FE) estimates of
the growth and equalization equations. This specification of the model does not account for the
potential simultaneous nature of growth and equalization or the potential endogeneity of fiscal
decentralization (FD) and FD-squared in the two equations. However, we believe that it is useful
to report and discuss these estimates to show that our main results are robust to alternative
specifications and not merely reflective of some kind of econometric chicanery.
The FE estimates are reported in the first columns of the two columnar-panels in Table 6,
labeled growth equation and equalization equation. The Hausman test leads us to favor a fixed
effect specification over random effects. Starting with the growth equation, the estimated
coefficient of equalization (E) is positive but statistically insignificant at conventional levels. The
estimated coefficient of fiscal decentralization (FD) is equal to -1.48 and is nearly four times its
30
standard error (SE = 0.36), and the estimated coefficient of FD-squared is 0.07 (SE = 0.02) and
statistically significant at conventional levels. The combined effect of FD and FD-squared has a
negative effect on growth for values of FD less than 21, and a positive effect for values greater
than 21. The effect of FD reaches a minimum at a value of 10.6. Since the average value of FD
in our sample is equal to 3.34, the average level of fiscal decentralization during the period
spanned by our data reduces the growth rate of GRP per capita by 4.2 percent which is
substantial in economic terms. In the equalization equation, the estimated coefficients of G and
FD are positive but statistically insignificant at conventional levels. However, FD-squared is
negative and statistically significant at conventional levels.
Although these results are certainly suggestive, the FE estimates, as previously noted,
may be biased and inconsistent due to the potential endogeneity of FD, FD-squared, E, and G.
Therefore, we estimate a simultaneous equation model (SEM) using two-stage least squares with
fixed effects (2SLS FE).
To identify these two equations, the two control sets cannot contain exactly the same set
of regressors; there must be valid exclusion restrictions in both equations. According to Jeffrey
M. Wooldridge (2002), the order condition for identification is satisfied for an equation in any
simultaneous equation model (SEM) if the excluded exogenous variables in the equation are not
less than the number of right-hand-side endogenous variables. The rank condition further
requires that at least one of the exogenous variables excluded from equation (1) must have a
nonzero coefficient in equation (2). As discussed in greater detail below, the rank condition is
satisfied for our SEM. The identification of equation (2) is the mirror image of that for equation
(1).
31
To implement 2SLS, we need instrumental variables for our potentially endogenous
variables. To be valid, the proposed instrumental variables must be correlated with the
potentially endogenous variables but uncorrelated with the error term. In the growth equation,
the excluded variables are the share of missing women (MWS), the natural logarithm of
population (LNPOP), and the extra-budgetary factor (XBGT). These variables serve as
instruments for the potentially endogenous variables in the growth equation. The share of
missing women (MWS) should influence the demand for equalization. Since the share of missing
women is unlikely to affect the growth rate, we believe that it is legitimate to exclude this
variable from the growth equation. We include the employment growth rate (L) in the growth
equation and exclude it from the equalization equation, and we include the natural logarithm of
the population (LPOP) in the equalization equation and exclude it from the growth equation.
Christine Wong (1998) points out that, extra-budgetary funds (XBGT) are counter-
equalizing, so XBGT should be correlated with E. To address the potential endogeneity of FD-
squared, we use lagged values, higher moments, and interactions of the instruments as additional
instruments. In the equalization equation, one-lag of G, one- and two-lags of FD, and squared
values of these variables are used as instruments. We believe that our identification strategy is
justified by economic and econometric theory. To address any remaining concerns about our
identification strategy, we also subject these specifications to a battery of statistical tests to verify
that our instruments are indeed valid.
The 2SLS FE estimates of the growth equation are reported in the second column of
Table 6. The estimated coefficient of equalization (E) has a positive but statistically insignificant
effect on the growth rate of real GRP per capita. The estimated coefficient of fiscal
decentralization (FD) is equal to -1.79 which is more than twice its standard error (SE = 0.74),
32
and the estimated coefficient of FD-squared is 0.19 (SE = 0.04) and statistically significant at
conventional levels. The non-linear effect of fiscal decentralization on growth is illustrated in
Figure 3. The negative effect of FD on growth reaches a minimum of -19 for a value of FD equal
to approximately 10, meaning that when the index of fiscal decentralization is equal to 10, the
growth rate decreases by 19 percent. As illustrated in Figure 3, fiscal decentralization has a
negative effect on economic growth for values of FD less than 20 and a positive effect for values
greater than 20.
As previously noted, the average degree of fiscal decentralization in China during the
period spanned by our data is substantially less than 10. The mean value of fiscal
decentralization in our sample is about 3.34, and this degree of FD reduces the growth rate by
approximately 10 percent. That is to say, this degree of fiscal decentralization currently
prevailing in China does not allow it to realize the potential positive effects of FD on economic
growth. This is consistent with the findings of a negative reported in Tao Zhang and Heng-Fu
Zou (1998) and inconsistent with the positive effect reported in Justin Yifu Lin and Zhiqiang Liu
(2000); Ying Ding (2007); and Qiao et al. (2008). In contrast to Tao Zhang and Heng-Fu Zou’s
findings, however, FD has a non-linear effect on growth in our model.
The estimated coefficients of many of the control variables in the growth equation
generally have the expected signs and are statistically significant at conventional levels. The
estimated coefficients of the following variables are positive and statistically significant at
conventional levels: the growth rate of investment in fixed assets (K), openness to trade (F), the
TSS reform dummy variable (TSS), central government development spending (CDEV),
government size (GOVS), rate of inflation (R), and the natural logarithm of foreign direct
33
investment (FDI). The estimated coefficient of the share of agriculture in GRP is negative and
statistically significant at conventional levels, as expected.
Turning now to the 2SLS FE estimates of the equalization equation reported in the
second column of the right-hand-side columnar-panel, we find that economic growth (G) has a
positive effect on equalization but is statistically indistinguishable from zero at conventional
levels. Similarly, fiscal decentralization is positive but statistically insignificant at conventional
levels. FD-squared, however, has a negative and statistically significant effect on equalization.
The estimated coefficients of the control variables have the anticipated signs and many
are statistically significant. More specifically, agricultural share (AGS), the TSS reforms (TSS),
central government development spending (CDEV), and the natural logarithm of population
(LNPOP) have a positive effect on equalization. Meanwhile, openness to trade (F), the share of
provincial revenues in GRDP (AET), government size (GOVS), extra-budgetary funds (XBGT),
and the share of missing women (MWS) have a negative effect on equalization. Regarding the
effect of extra-budgetary funds (XBGT), our findings are consistent with the findings of
Christine Wong (1998) and the World Bank (2000, 2001), and World Bank (2001) that extra-
budget funds are contributing to interregional fiscal disparities. Perhaps the most gratifying
finding is that the 1994 TSS reforms appear to be increasing growth and equalization which were
major goals of the reforms.
Sometimes, 2SLS FE can be problematic and even worse than ordinary least squares.16
Since the number of IVs is greater than the number of potentially endogenous variables, the
system is over-identified. Lars Peter Hansen (1982) suggests GMM as a solution. Two-step
GMM, with IVs to deal with over-identification, allows for arbitrary autocorrelation and cross-
16 John Bound, David Jaeger, and Regina Baker (1996) point out that if the instruments are weakly correlated with endogenous variables, IV estimation could be worse than OLS.
34
sectional heteroskedasticity. The estimated coefficients of the two-step GMM specification of
our model are provided in the third columns of the two-columnar panels in Table 6. However,
these results are nearly identical to those for FE and 2SLS FE.
We test for the potential endogeneity of fiscal decentralization (FD), FD-squared,
horizontal fiscal decentralization (E), and growth (G) and conduct a number of other
specification tests. Table 7 summarizes the results of this battery of tests. Following Jerry A.
Hausman (1978), we estimate the Hausman test of exogeneity. We fail to reject the joint null
hypothesis of exogeneity for our set of potentially endogenous variables (E, G, FD, FD-squared)
in both equations. We further estimate a Durbin-Wu-Hausman chi-squared test statistic for the
null hypothesis that a given variable is exogenous independently of the other potentially
endogenous variables. The small values of the test statistics for FD, and E in the growth equation
lead us to fail to reject the exogeneity of these two variables, one independently of the other.
However, we reject the null hypothesis that FD-squared is exogenous at the 10 percent
significance level. In the equalization equation, the chi-squared statistic for G is too small to
reject the null hypothesis of exogeneity. However, we soundly reject the exogeneity of FD and
FD-squared in the equalization equation.
Since endogeneity may go undetected due to weak instruments, James H. Stock and
Motohiro Yogo (2005) provide a rule-of-thumb test for weak instruments and critical values for
this statistic. The null hypothesis for this test is that all IVs are sufficiently correlated with the
potentially endogenous variables that the small sample bias of 2SLS is less than the bias of FE.
Using the Cragg-Donald Wald F values, the weak identification tests show that the bias of the IV
estimate of the growth equation is less than 20 percent that of the OLS estimates and less than 5
percent of OLS estimates in the equalization equation. In other words, the set of instruments are
35
sufficiently, strongly correlated with the potentially endogenous variables that the small sample
bias of 2SLS is small relative to that of OLS.
Finally, John D. Sargan (1958), Lars Peter Hansen (1982), and Lars Peter Hansen and
Kenneth J. Singleton (1982) show in a GMM context that when the number of moment
conditions exceeds the number of parameters to be estimated, a chi-square test can be used to
determine whether the instruments are orthogonal to the error term. The values of the Sargan
statistics for the test of over identifying restrictions in the growth and equalization equations are
10.53 and 0.54, respectively. These values fail to reject the null hypothesis that the instruments
are exogenous.
To further test the robustness of these results, we use our data to conduct an “extreme-
bounds analysis” (EBA) for the growth equation, based on the work of Edward Leamer (1978,
1983, and 1985), Ross Levine and David Renelt (1992), and Tao Zhang and Heng-Fu Zou (1998).
In our base scenario, we include the variables of primary interest (I), specifically fiscal
decentralization (FD) and horizontal fiscal equalization (E), and the main model variables (M),
including capital (K), labor (L), openness (F), and inflation rate (R). We proceed by estimating
thousands of variations of the base scenario described above, by adding every combination of the
remaining 16 variables in sets including 0 to four (4) optional variables (O). Then, we identify
the control set of optional variables that give us minimum and maximum values of the estimated
coefficients of each of the primary variables of interests and the main variables.
Table 8 summarizes the results of our extreme bounds analysis. Among the set of
interesting (I) and main (M) variables, only the estimated coefficients of K and L are same
signed and more than two standard deviations from zero for all combinations up to 4 of the
optional variables. Therefore, capital and labor are the only robust variables among all the
36
growth determinants considered in this study. The base and lower bound estimates for fiscal
decentralization (FD) are negative while the upper bound is positive. The extreme bounds for
equalization (E) cannot be found. Therefore, these two variables are fragile, as are the other main
models variables. Based on these findings, we are somewhat assured that it is appropriate to
include fiscal decentralization, equalization, capital, labor, openness, and the rate of inflation in
the growth equation.
The reason that the results discussed above differ from those of Qiao et al. (2008) may be
due to the different time periods employed in the two studies. To explore this issue, we examine
the robustness of our results by replicating Qiao et al. (2008), using their variable definitions and
model specifications and different time periods of our data. Because extra-budget data are not
available for the years before 1985 or after 2006, the definitions of fiscal decentralization and
equalization include provincial budget expenditures only.
The regression results of these experiments are summarized in Table 9. For ease of
comparison, we report the estimated results reported in Qiao et al. (2008) for ease of comparison
in the first columnar-panel. In the second columnar-panel, we report the estimated coefficients
based on their specification and time period but using our data which differ slightly from theirs.
In the third columnar-panel, we report the estimated coefficients based on their specification, our
time period (1985-2006), and our data. Finally, in the fourth columnar-panel, we report the
estimated coefficients from estimating their specification using an expanded time period (1949-
2006), and our data. The finding of these experiments is that our results appear to be robust to
their specification and a variety of different time periods.
37
Conclusions
We find that fiscal decentralization has a non-linear effect on economic growth. For
values between 0 and 20, fiscal decentralization has a negative effect on economic growth. In
this range, our finding of a negative effect is consistent with that of Tao Zhang and Heng-Fu Zou
(1998), but inconsistent with the findings of a positive effect of Justin Yifu Lin and Zhiqiang Liu
(2000); Ying Ding (2005); and Qiao et al. (2008). Beyond the value of 20, fiscal decentralization
has a positive effect on economic growth. Qiao et al. (2008) report evidence of a trade-off
between equalization and growth. In contrast, we find evidence of a weak positive relationship
between equalization and growth.
Based on these results, it would appear that a significant deepening of fiscal
decentralization in China would have a beneficial effect on the performance of the economy in
terms of equalization and growth. A higher growth rate may have a weak, positive effect on
equalization but certainly does not lead to greater interregional fiscal disparities, which appears
to be the concern of many students of China’s local public finances. Finally, equalization does
not appear to influence the growth rate. Finally, we find that the 1994 TSS reforms are having a
positive effect on the performance of China’s economy, both in terms of growth and equalization.
38
Chapter III: Examining the Potential Trade-off between
Horizontal Fiscal Equalization and Growth in India
Introduction
Following the economic reforms of the early 1990s, India has experienced an
acceleration of economic growth and a reduction in its poverty rate. During the past five years,
the annual growth rate of real gross domestic product (GDP) has averaged almost nine percent,
in contrast with a four percent growth rate during the period from 1950 to 1989. As result of this
recent period of extraordinary growth, the poorest members of Indian society have benefitted:
almost 30 million people escaped poverty during the period 1980 to 2005. This accounts for
nearly 4.5 percent of the global population in poverty, based on the dollar-per-day poverty
standard used by the World Bank.17
However, this impressive record of growth has also increased income inequality among
the states. Specifically, in 1980 Goa was the richest state, with a real gross state domestic
product (GSDP) of $252 per capita.18 In contrast, the poorest state was Bihar, with a real GSDP
of $46 per capita. In other words, Goa’s real GSDP per capita was five times that of Bihar’s.
Interregional income inequality remained stable until the outset of the 1992 economic reforms.
By 2005, Goa’s real GSDP per capita increased from five times that of Bihar’s to nearly eight
times, or $1,230 versus $157 per capita, respectively.
With the recent upward revision in its poverty norm from one dollar-per-day to one and a
quarter dollars per day, the number of people living in poverty in India’s increased by 35.3
million, according to a World Bank study conducted by Shaohua Chen and Martin Ravallion
17 See the World Development Report from 1990 for information about the poverty standard. 18 We use the INR-U.S. dollar exchange rate, which was 0.0211797 on May 24, 2009 (http://www.xe.com/ucc/convert.cgi).
39
(2008). However, the percentage of people living on less than this new poverty line has come
down from 60 in 1981 to 42 in 2005.
Besides increased per capita income inequality, inequality in per capita expenditures by
sub-national governments has also increased. For example, in 1981 Nagaland had the highest
government expenditures, with $36 per capita. The lowest state government per expenditures was
in Uttar Pradesh, with $6 per capita, or one-sixth that of Nagaland’s per capita expenditures. By
2005, the highest state government expenditures occurred in Sikkim, with $795 per capita, which
is more than 13 times greater than that of Bihar, the lowest, at $54 per capita. According to
Figure 4, the coefficient of variation (CV) of state total consolidated expenditures per capita was
64 percent in 1981, which increased to 80 percent in 2005. Several previous studies find that the
equalization effect of the intergovernmental transfer system in India is not performing well, at
least in terms of equalizing expenditures per capita. Govinda Rao and Nirvikar Singh (2000)
charge that the lack of coordination among the myriad of different transfers working at cross-
purposes to one another have led to an increase in fiscal disparities among the states. Roy Bahl,
Eunice Heredia-Ortiz, Jorge Martinez-Vazquez and Mark Rider (2005a) attribute the lack of an
adequate equalization transfer system to the lack of a clear link between fiscal capacity and
expenditure needs as well as lack of transparency and the less equalizing Gadgil formula. D. K.
Srivastava (2002) shows some variation in state fiscal capacities in terms of state per capita
expenditures on services with data for the period from 1996 to 1998.
However, C. Rangarajan and D. K. Srivastava (2008) find that despite different state
fiscal capacities, a high degree of horizontal fiscal equalization has been achieved through
recommended transfers by the Twelfth Finance Commission. Based on our data, the
intergovernmental transfer system in India has resulted in greater equalization in state per capita
40
expenditures since the beginning of the economic reforms in the early 1990s. For example, the
CV of state expenditures per capita excluding gross devolution was 89 percent in 2005 and
declined to 80 percent when taking into account gross devolution. As Figure 4 shows, gross
devolution began to have an equalizing effect on state per capita expenditures after 1994. Figure
4 also shows the trend in the CV for state per capita expenditures less grants, and the trend in the
CV for state per capita expenditures less grants and less state share of central taxes. These trends
clearly show that central government transfers have had an equalizing effect on state
expenditures per capita.
The question that arises from this analysis of India’s recent growth experience is whether
India’s policy of fiscal decentralization (FD) and horizontal fiscal equalization (HFE) has
encouraged or impeded economic growth. From a theoretical perspective, we contend that the
effects of FD and HFE on economic growth are ambiguous.
The reasons for an ambiguous relationship between HFE and economic growth are
similar to those for the case of China; therefore we will only briefly review some of the
arguments here. HFE involves shifting resources from high- to low-income regions, resulting in
less public infrastructure investments in high-income regions. However, the resulting decrease in
growth promoting public infrastructure investments in high-income regions may be more than
the increase in public infrastructure investments in low-regions because of the distortionary taxes
used to finance the transfer system. Furthermore, low-income regions may not use transfers as
efficiently as high-income regions. Even though, the investment resources are transferred
completely and may not be utilized in low-income regions as at least efficiently as in high-
income regions. Finally, equalizing transfers may be useful in reducing civil unrest in low-
41
income regions populated by minority groups or other aggrieved people. From a theoretical
perspective, equalizing transfers have an ambiguous effect on the aggregate national growth rate.
The empirical evidence is mixed, as well. Georgia Guess, William Loehr and Jorge
Martinez-Vazquez (1997), Elliott Parker and Judith Thornton (2007), and Baoyun Qiao, Jorge
Martinez-Vazquez and Yongsheng Xu (2008) find that there is a trade-off between HFE and
growth. The purpose of this study is to examine the effect of India’s policy of fiscal
decentralization and horizontal fiscal equalization on economic growth. Following the work of
Qiao et al. (2008), we use a simultaneous equations model (SEM). Using an instrumental
variables approach, this allows us to account for the potential endogeneity of fiscal
decentralization, the square of fiscal decentralization, horizontal fiscal equalization, and growth.
We apply a variety of estimation methods using state-wide data for India covering the period
from 1980 to 2005.
This study makes the following contributions to the existing literature on fiscal
decentralization and growth. First, there is no previous research that explicitly examines the
relationship between the twin policies of fiscal decentralization and horizontal fiscal equalization
on economic growth in India. This study attempts to fill this gap. Second, we use state-wide
yearly data from 1980 to 2005, which straddles the market and decentralization reforms initiated
in the early 1990s. This panel data set enables us to more thoroughly evaluate the performance of
economic reforms in the early 1990s. Third, we employ a complete measure of fiscal
decentralization and horizontal fiscal equalization, by accounting for both revenue expenditures
and capital expenditures. Fourth, we account for the potential endogeneity of fiscal
decentralization, the square of fiscal decentralization, horizontal fiscal equalization, and
economic growth in our two equation model over time and among the states. Finally, we exploit
42
the panel nature of our data by estimating ordinary fixed effects (FE), two-stage least squares
fixed effects (2SLS FE), and two-step generalized method of moments (GMM) models.
The remainder of this chapter is organized as follows. The next section provides some
background on the legal and institutional aspects of India’s intergovernmental fiscal system. In
the third section, we summarize the previous literature on fiscal decentralization in India and
some issues in measuring in fiscal decentralization. The fourth section describes the econometric
model, variable construction, and data. Section five discusses our empirical results, and the final
section concludes.
India’s Intergovernmental Fiscal System
India is the most populous democracy in the world. It has a parliamentary system in
which the prime minister, as the head of the central government and the cabinet are chosen by
the party or party coalition that wins a majority of the votes. India is a federal republic
comprising one central government (the Union), twenty-eight states, and seven union territories
(the States), including the National Capital Territory of Delhi. Since its foundation in 1947 up to
the early 1990s, India has been a two-tier federation with a constitutional demarcation of
responsibilities and finances between the Union and the States, which have separate legislative,
executive, and judicial organs of governing. Fiscal decentralization never went beyond the state
level until the enactment of the Constitutional amendments in 1992, granting statutory
recognition to rural and urban local self governments, although these local bodies long pre-dated
independence in 1947. There are about 610 districts (Zila Parishads) in total, which are further
divided into about 6,000 blocks (Panchayat Samiti) and 250,000 villages (Gram Panchayats) in
rural areas. Municipal corporations, municipalities, and councils constitute the third-tier of urban
43
governments.19 Compared to the leader-subordinate relationship in the center-provincial
framework of China, both India’s executive officer of the central government and that of the
state government are accountable first to the corresponding legislature and second to the voters,
though a guiding administrative relationship exists between the Union and the states. It is the
legislative power that dominates in the institutional system.
At the same time, India has a long history of local self-governance, through the gradual
evolution of decentralization, which can be dated back to Lord Mayo’s Resolution in 1870 with
the introduction of an elected president in the municipalities. However, local self governments
only gained constitutional status and federal recognition with the 73rd and 74th amendments to
the Constitution in 1992, granting greater powers and fiscal resources to rural and urban local
self governments, respectively. The 11th Schedule lists 29 items as the responsibilities of rural
bodies, and the 12th Schedule lists 18 items as responsibilities of urban local bodies. Govinda
Rao and Nirvikar Singh (1999) point out several institutional failures with India’s Constitutional
arrangements, including a clear centripetal bias in the distribution of fiscal powers, the separation
principle of taxing powers as opposed to concurrent tax assignments, concurrent expenditure
assignments rather than exclusive expenditure assignments, and constitutional permission to tax
inter-state trade that hinder development of an effective intergovernmental fiscal system in India.
Before the 1980s, fiscal centralization dominated the national economy. Beginning in the
mid-1980s, fiscal decentralization was the focus for revitalizing the economy. Fiscal
decentralization, together with market reforms, brought about high fiscal deficits as well as a
high economic growth rate. The combined gross fiscal deficit of the Union and state
governments as a percentage of GDP increased from 7.4 in 1980 to 9.4 in 1990. The combined
19 Data is retrieved in December, 2008 from http://districts.gov.in/, government of India. National Informatics Centre (NIC).
44
outstanding liabilities of the central and state governments as a percentage of GDP increased
from 54 in 1984 to 67 in 1990. The economic reforms initiated in the 1980s did not change the
fundamental framework of the fiscal policy or development policy. Driven by the debt crisis in
the early 1990s, the Indian government began a process of deregulation, privatization, and
liberalization of the economy, together with enlarging the tax bases, compressing expenditures,
and strengthening state fiscal discipline with the enactment of the Fiscal Responsibility and
Budget Management Bill in 2000. The new round of economic reforms has resulted in lower
fiscal deficits as well as a remarkable increase in the annual economic growth. As a result, the
combined central and state gross fiscal deficit fell to 6.7 percent of the GDP in 2005, although
the high outstanding liabilities remain. This chapter examines the influence of India’s policy of
fiscal decentralization and horizontal fiscal equalization on economic growth.
We turn now to a description of India’s four pillars of fiscal decentralization. First,
India’s expenditure assignments, as provided in Article 246 of the India Constitution (also
known as the Seventh Schedule), have three distinct lists: the Union list (List 1) specifies those
functions which are the exclusive competence of the Union government; State list (List 2: State
List) specifies those functions which are the exclusive competence of the state governments; and
the Concurrent List (List 3) specifies those functions that are the current responsibility of Union
and state governments. Table 10 provides a summary of India’s statutory expenditure
assignments. The murkiness lies in the current responsibilities in List III in which the role of
states is unclear. Due to the lack of accountability resulting from the extensive use of concurrent
expenditure responsibilities, a large share of state revenues are spent on relatively unproductive
expenditures, such as administration, debt services, and pensions. Non-development
expenditures as a share of total recurrent expenditures increased from less than 19 percent in
45
1980 to 34 percent in 2005. In the meantime, over 60 percent of agricultural services and nearly
90 percent of educational, public health, family welfare, electricity, and irrigation expenditures
are shouldered by state governments.
Table 11 summarizes revenue assignments as provided by the three lists in the Seventh
Schedule. The most productive taxes are exclusively assigned to the Union government. At the
same time, the states have been assigned a number of tax handles. However, from the viewpoint
of productivity and buoyancy, only the sales and purchase taxes are important. Since the 80th
Amendment to the Constitution in 2000, instituted to improve state finances, the net proceeds of
all Union taxes and duties are now shared taxes between the Union and the States. Furthermore,
the 88th Amendment to the Constitution in 2003 includes taxes on services in List I as shared
taxes with the states. Figure 5 shows the trends in Union and state shares of total revenues and
expenditures. The proportion of all fiscal revenues shared by the Union and States changed from
60:40 before 2000 to 50:50 in 2005. In addition, although 29 expenditure items under the 73rd
amendment and 18 expenditure items under the 74th amendment have been assigned to rural
Panchayats and urban municipalities, respectively, neither of them has sufficient revenue
autonomy. For the third-tier government in rural areas, only Gram Pachayat has independent
revenues, such as non-agricultural property tax, title tax, water usage fees, and others. However,
Melville L. McMillan (2007) estimates that these own source revenues account for less than one-
quarter of total local revenue for the State of Karnataka which is relatively advanced in terms of
development and decentralization to the local level.
The vertical and horizontal imbalances created by the revenue and expenditure
assignments are well known. The Constitution provides for an equalization grant, and the Union
government administers a large number of centrally sponsored schemes (CSS) to help finance
46
state and local expenditure responsibilities. There are several channels through which
intergovernmental transfers are received by sub-national governments. The first channel is the
tax share in centrally levied taxes, which are determined and transferred by the Finance
Commission of the Union. The recommendations and transfer formulas by the Finance
Commission rely heavily on per capita income to measure the backwardness of the recipient
states. The second channel is the Central Planning Commission’s capital grants and loans. In
general, these plan grants and loans were project oriented, but they were later based on the so-
called Gadgil formula, which has been modified several times with a proportionate mix of grants
and loans. These grants and loans can be roughly classified into two categories: central
assistance and additional central assistance. The third channel is through central ministries and
departments, under which a broad range of development programs known as Centrally
Sponsored Schemes (CSSs) are initiated by the Union and implemented by various departments
of the states. The number of CSS types is estimated to number well over two hundred and cover
a variety of development tasks like poverty alleviation, family planning, irrigation, and education.
Among the three main components of gross devolution, shares in central taxes and central
grants follow a similar increasing trend, while loans from the Union have shrunk significantly
since 2003 to less than 10 percent of total gross devolution. This is due to the desire of the Union
government to impose fiscal discipline on state governments. Grants from the Union, non-plan
schemes, centrally sponsored schemes, and state plan schemes account for a more than 90
percent of total revenues in recent years. The multiplicity of transfer channels, together with the
lack of coordination between the Finance Commission and the Planning Commission,
undermines the fiscal autonomy of sub-national governments and may foster a lack of fiscal
discipline due to the high level of transfer dependency. Additionally, Roy Bahl et al. (2005a),
47
and Eunice Heredia-Ortiz and Mark Rider (2005) point out several shortcomings of the Planning
Commission’s transfers. Although the transfer system in India has received much criticism, the
transfers have decreased the CV of state per capita expenditures, as shown in Figure 4.
Finally, sub-national borrowings are allowed in India with certain restrictions, such as no
access to direct foreign borrowings. Besides borrowings in the form of plan loans from the Union,
domestic bonds and policy loans from international organizations provide market loans. Since
2003, central loans have decreased, and states have resorted to more market borrowings, which
involve heavy interest payments by the state.
As discussed in greater detail in Roy Bahl et al. (2005a), India’s intergovernmental fiscal
system is characterized, much like China’s system, by an over centralization of revenues, and
over decentralization of expenditures, high level of transfer dependency, and excessive sub-
national borrowing.
Literature Review
To the best of our knowledge there are no rigorous empirical studies of the type
undertaken here. However, there is an enormous literature on fiscal decentralization in India.
Govinda Rao and Nirvikar Singh (1999) contend that fiscal decentralization can give rise to an
unhealthy “race to the bottom” among sub-national governments competing to attract mobile
factors of production, particularly capital. In addition, such competition can hinder the
development of a common market, particularly when there are no Constitution prohibitions
against taxes on inter-state trade. Catriona Purfield (2004) concludes that institutional factors
play a key role in the differing fiscal performances among the states of India. In his study, state
fiscal performance is mainly measured by fiscal deficits and debts. He uses panel data from 1985
48
to 2000 for the 15 largest states and finds a significantly positive relationship between fiscal
decentralization and the size of the state fiscal deficit.
Govinda Rao (2003) conducts a comparative analysis of fiscal decentralization in China
and India. He identifies the emerging challenges and puts forward a solution or overall approach.
Jorge Martinez-Vazquez and Mark Rider (2006) link fiscal decentralization to economic growth
in China and India and conclude that poor service delivery and poor fiscal self-discipline at the
sub-national level threatens the ability of both countries to sustain high rates of economic growth
over the long-run.
As previously discussed, there is no single measure of fiscal decentralization. Like China,
India has chosen to centralize revenues and decentralize expenditures. Therefore, we use
expenditure decentralization as an index of fiscal decentralization among the states of India over
time. Accordingly, our index of fiscal decentralization (FD) is the ratio of state expenditures to
central government expenditures in per capita terms. This index includes both revenue
expenditures and capital expenditures at the central and state levels. This measure is used in a
number of well-known studies of fiscal decentralization.20
Empirical Strategy
In this section, we describe our empirical model, variable construction, and the data used
in this analysis. As previously noted the effect of fiscal decentralization and horizontal fiscal
equalization on economic growth is ambiguous. To gauge this relationship, we estimate a
simultaneous equation model (SEM), consisting of two equations: growth equation and an
equalization equation. This model is very similar to examine the case of China; therefore, we
refer the interested reader to Chapter II for further details.
20 Kunrong Shen and Wenlin Fu (2005), Andrew Feltenstein and Shigeru Iwata (2005), Nobuo Akai and Masayo Sakata (2002), Tao Zhang and Hengfu Zou (1998), etc.
49
Following Qiao et al. (2008), we estimate the following SEM:
itiitititit
itititititit
ERCATDSTCT
L�POPKFDFDEG
υµαααα
αααααα
+++++
++++++=
8776
54
2
3210
(3)
itiititititit
ititititit
ERCATDMWSFDRFSR
L�POPFDFDGE
εδβββββ
βββββ
++++++
+++++=
98765
4
2
3210
(4)
The variable definitions are as follows. The subscript i is the observation unit, province; t
is the time unit, year. G is the growth rate in real GRP per capita. E is an index of equalization
and follows the definition of Qiao et al. (2008). More specifically, we compute the difference in
year t between a province’s per capita expenditures in year t and the mean of provincial
expenditures in year t, for every province and year in our sample. We convert these differences
into ‘distances from the mean per capita provincial expenditures in year t’, by taking the absolute
value of these figures. Then, we normalize these year t ‘distances’ by dividing each of them by
the year t mean per capita provincial expenditures. Finally, we apply a minus sign to these
figures in order to give our index of equalization (E) a more intuitive interpretation. As E
approaches zero from the left, the province’s per capita expenditures are becoming closer to the
mean of per capita provincial expenditures and thus becoming more equal.
Other control variables in the growth equation include capital (K) measured as the natural
logarithm of capital outlays per capita, and labor (L) measured as the natural logarithm of the
state population due to the unavailability of actual labor force data. According to neo-classical
growth models like that of Robert J. Barro (1990), both of these variables are expected to have a
positive impact on G. In addition, the Union governments average effective tax rate (CT) defined
as central revenue receipts as a share of GDP, and state average effective tax rate (ST), defined
as state revenue receipts as a share of GSDP, are included to control the effect of tax effort. Since
these are distortionary taxes, they are expected to have a negative effect on growth.
50
Jean-Philippe Meloche, Francois Vaillancourt and Serdar Yilmaz (2004) contend that the
rate of fiscal self-reliance (FSR) and rate of transfer dependence (FDR) are very important
determinants of the horizontal fiscal balance among sub-national governments. Therefore, these
two variables are included in the equalization equation. FSR is the share of state’s own revenues
in state revenue receipts, and FDR is the share of central government grants in total state revenue.
Furthermore, we include the share of missing women (MWS) in the total state population, which
is assumed to have an adverse effect on horizontal equalization. For example, increased pensions
are required for elderly men who never married because of a biased sex ratio due to the strong
cultural preference for a male child.21 Based on the method of Ansley J. Coale and Judith
Banister (1994), we calculate the number of hypothesized missing women for each state
according to the hypothetical sex ratio in Amartya Sen (1990).
CATD is a dummy variable set equal to “1” in the year and state in which there was a
natural catastrophe that one would expect directly to influence agricultural production and
regional fiscal capacity. ER is a time dummy variable set equal to “1” for years after 1991 and
“0” otherwise. This variable controls for the economic reforms that came into effect in 1992. We
pursue the same identification strategy as before and refer the interested reader to the discussion
in Chapter II.
The terms µi and δi are unobserved regional effects, which are time-invariant. The terms
υit and εit are idiosyncratic error terms, which are time-varying and represent unobserved factors
that change over time and influence the dependent variable.
The data used in this study are state-wide for the years 1980 through 2005. The main
source for India’s state fiscal and general national accounts data is the Reserve Bank of India
21 See, for example, Quanbao Jiang, Marcus W. Feldman and Xiaoyi Jin (2005) for the adverse effects of missing women.
51
(retrieved in December, 2009 from www.rbi.org.in). All-India data are from the Central
Statistical Organization, and population data are from India’s Office of the Registrar General.
Data on natural catastrophes are from Natural Disaster Management, Ministry of Home Affairs.
All real GSDP per capita figures have been adjusted to 1993–1994 base year.
The following union territories are excluded from the sample: Andaman and Nicobar
Islands, Chandigarh, Dadra and Nagar Haveli, Daman and Diu, Lakshadweep, and Puducherry.
Thus, there are a total of 28 state governments and the national capital territory of Delhi, making
for 29 states or comparable units. These 29 state governments include the States of Chhattisgarh,
Jharkhand, and Uttaranchal which were established through the bifurcation of the States of
Andra Pradesh, Bihar, and Uttar Pradesh in 2000. Therefore, the data for Chhattisgarh,
Uttaranchal, and Jharkhand are available only for the 6 year period from 2000 to 2005. The
resulting sample would consist of 694 observations (26 × 26 + 3 × 6 = 694). In calculating state
growth rate in real GSDP per capita, we lose the first year of our data or 29 observations. In
addition, complete data for Mizoram are only available for the period beginning in 2000, and
complete data for Arunachal Pradesh and Goa are available for the period beginning in 1986.
Complete data for Nagaland, Sikkim and Delhi is available beginning in 1994. Consequently,
there are 562 (= 694 - 2×29 - 1×20 - 2×6 - 3×14) observations in the sample used to estimate the
FE specifications. For the 2SLS FE and GMM specifications, the number of observations is
somewhat less than that for the FE specifications because of missing values for some of the
instrumental variables.
Table 12 provides a summary of the variable definitions, and Table 13 provides summary
statistics for our sample of 562 observations used to estimate the FE specifications. The average
growth rate in real GSDP per capita (G) in our sample is 4.67 percent, the maximum growth rate
52
is 81.67 percent, and the minimum is -43.7 percent. The CV of the provincial growth rates in our
sample is 292.3 percent, meaning that the standard deviation of the growth rates in our sample is
nearly 300 percent of the mean growth rate. The degree of fiscal decentralization (FD) in our
sample ranges from 0.27 to 12.18, with a mean value of 1.65. Horizontal fiscal equalization (E)
varies between -4.78 and 0.0, with a mean value of -0.51. The average of the natural logarithm of
capital outlays per capita (K) is 5.55, and the average of the natural logarithm of state population
(LNPOP), which serves as a proxy variable for the size of the labor force, is 16.64. The average
share of central revenue receipts in GDP (CT) is 9 percent, and the average share of state revenue
receipts in GSDP (ST) is 25 percent. The average share of state own tax and non-tax revenues in
total revenue receipts (FSR) is 50 percent and the average share of central grants in consolidated
receipts (FDR) is 29 percent. Finally, the average share of missing women (MWS) is 3 percent of
the state population, and varies between 0.3 and 6 percent in our sample. Now, we turn to a
discussion of our empirical results.
Empirical Results
This section begins by reporting the estimation results for a variety of specifications.
Second, we carry out detailed specification tests. Third, we conduct an “extreme bounds
analysis” (EBA) for the growth equation to justify our selection of the baseline model reported in
Table 14.
Table 14 provides the main results for our baseline specifications. We begin by
discussing the fixed effects (FE) estimates of the growth and equalization equations. This
specification of the model does not account for the potential simultaneous nature of growth and
equalization or the potential endogeneity of fiscal decentralization (FD) and FD-squared in the
53
two equations. However, we believe that it is useful to report and discuss these estimates to show
that our main results are robust to alternative specifications.
The FE estimates are reported in the first columns of the two columnar-panels in Table
14, labeled growth equation and equalization equation. The Hausman test leads us to favor a
fixed effect specification rather than a random effects specification. Starting with the growth
equation, the estimated coefficient of equalization (E) is positive but statistically insignificant at
conventional levels. The estimated coefficients for both fiscal decentralization (FD) and FD-
squared are positive but statistically insignificant at conventional levels. In the equalization
equation, the estimated coefficient of G is positive and statistically significant at the 10 percent
level. This means that growth (G) has a positive effect on equalization (E), rather than growth
exacerbating fiscal disparities among the states. As for the effects of fiscal decentralization, both
FD and FD-squared are negative and statistically significant at conventional levels.
Although these results are certainly suggestive, the FE estimates, as previously noted,
may be biased and inconsistent due to the potential endogeneity of FD, FD-squared, E, and G.
Therefore, we estimate a simultaneous equation model (SEM) using two-stage least squares with
fixed effects (2SLS FE). To identify these two equations, the two control sets cannot contain
exactly the same set of regressors; there must be valid exclusion restrictions in both equations.
According to Jeffrey M. Wooldridge (2002), the order condition for identification is satisfied for
an equation in any simultaneous equation model (SEM) if the excluded exogenous variables in
the equation are not less than the number of right-hand-side endogenous variables. The rank
condition further requires that at least one of the exogenous variables excluded from equation (3)
must have a nonzero coefficient in equation (4). As discussed in greater detail below, the rank
54
condition is satisfied for our SEM. The identification of equation (4) is the mirror image of that
for equation (3).
To implement 2SLS, we need instrumental variables for our potentially endogenous
variables. To be valid, the proposed instrumental variables must be correlated with the
potentially endogenous variables but uncorrelated with the error term. In the growth equation,
the excluded variable: the share of missing women (MWS) is used to serve as instrument for the
potentially endogenous variables in the growth equation. The share of missing women (MWS)
should influence the demand for equalization, but it is unlikely to affect the growth rate.
Therefore, we conclude that it is appropriate to exclude this variable from the growth equation.
To address the potential endogeneity of FD-squared, we use lagged values, higher
moments, and interactions of the instruments as additional instruments. In the growth equation,
one-lag of E, one-lag of FD, interacted values, squared values, and cubic values of these
variables are used as instruments. In the equalization equation, one-lag of G, one-lag of FD, and
squared values of these variables are used as instruments. We believe that our identification
strategy is justified by economic and econometric theory. To address any remaining concerns
about our identification strategy, we also subject these specifications to a battery of statistical
tests to verify that our instruments are indeed valid.
The 2SLS FE estimates of the growth equation are reported in the second column of
Table 14. The estimated coefficient of equalization (E) is negative but statistically insignificant
effect at conventional levels. The estimated coefficient of fiscal decentralization (FD) is equal to
-62.38, which is about five times its standard error (SE = 12.71), and the estimated coefficient of
FD-squared is 3.05 (SE = 1.04) and statistically significant at conventional levels. The non-linear
effect of fiscal decentralization on growth is illustrated in Figure 6. The negative effect of FD on
55
growth reaches a minimum of -319 for a value of the FD index equal to approximately 10.
Meaning that when the index of fiscal decentralization is equal to 10, the growth rate decreases
by 319 percent. As illustrated in Figure 6, fiscal decentralization has a negative effect on
economic growth for values of FD less than 21 and a positive effect for values greater than 21.
As previously noted, the average degree of fiscal decentralization in India during the
period spanned by our data is substantially less than 10. The mean value of fiscal
decentralization in our sample is about 1.65, and this degree of FD reduces the growth rate by
approximately 100 percent. That is to say, this degree of fiscal decentralization currently
prevailing in India does not allow it to realize the potential positive effects of FD on economic
growth. This is consistent with our findings for China. It is also noteworthy that the minimum
point and the point where the effect of FD switches from a negative effect to a positive effect are
remarkably similar for these two countries.
The estimated coefficients of the other control variables in the growth equation generally
have the expected signs. The natural logarithm of capital outlays per capita (K) is positive and
statistically significant at the 1 percent level. The central revenue share in GDP (CT) is negative
and the state revenue share in GSDP (ST) is positive, but both are statistically insignificant at
conventional levels. The estimated coefficient of the economic reform dummy variable (ER) is
positive but statistically insignificant at conventional levels. The natural logarithm of state
population (LNPOP) is negative and statistically significant at conventional levels, which may
indicate that state population is a poor proxy for the size of the labor force.
Turning now to the 2SLS FE estimates of the equalization equation reported in the
second column of the right-hand-side columnar-panel, we find that economic growth (G) has a
positive effect on equalization and is statistically significant at conventional levels. The
56
estimated coefficient of growth (G) is equal to 0.008 and is more than twice its standard error
(SE = 0.003). This estimate implies that a percentage increase in the growth rate of real GSDP
per capita increases equalization by 0.01. Fiscal decentralization (FD) is positive and statistically
significant at conventional levels in contrast to the negative effect obtained with the FE
specification. The possible reason being FD is endogenous and therefore the FE estimate is not
consistent. FD-squared is negative and statistically significant at conventional levels.
The estimated coefficients of the other control variables are as following. Missing women
(MWS) has the anticipated negative sign and is statistically significant. The natural logarithm of
population (LNPOP) has a positive effect on equalization and is statistically significant at
conventional level. Meanwhile, the rate of grant dependence rate (FDR) is positive but
statistically insignificant at conventional levels. The rate of fiscal self-reliance rate (FSR) is
negative and statistically significant at conventional levels. The economic reform time dummy
variable (ER) is negative sign but statistically indistinguishable from zero at conventional levels.
Sometimes, 2SLS FE can be problematic and even worse than ordinary least squares.22
Since the number of IVs is greater than the number of potentially endogenous variables, the
system is over-identified. Lars Peter Hansen (1982) suggests GMM as a solution. Two-step
GMM, with IVs to deal with over-identification, allows for arbitrary autocorrelation and cross-
sectional heteroskedasticity. The estimated coefficients of the two-step GMM specification of
our model are provided in the third columns of the two-columnar panels in Table 14. However,
these results are nearly identical to those obtained with the 2SLS FE specification.
We test for the potential endogeneity of fiscal decentralization (FD), FD-squared,
horizontal fiscal decentralization (E), and growth (G) and conduct a number of other
22 John Bound, David Jaeger, and Regina Baker (1996) point out that if the instruments are weakly correlated with endogenous variables, IV estimation could be worse than OLS.
57
specification tests. Table 15 summarizes the results of this battery of tests. Following Jerry A.
Hausman (1978), we estimate the Hausman test of exogeneity. We fail to reject the joint null
hypothesis of exogeneity for our set of potentially endogenous variables in the equalization
equation while soundly rejecting the joint null hypothesis of exogeneity in the growth equation.
We further estimate a Durbin-Wu-Hausman chi-squared test statistic for the null hypothesis that
a given variable is exogenous independently of the other potentially endogenous variables. The
large values of the test statistics for FD, FD-squared and E in the growth equation lead us to
reject the exogeneity of these three variables again. In the equalization equation, the chi-squared
statistic for FD-squared is too small to reject the null hypothesis of exogeneity. However, we
soundly reject the exogeneity of FD and G in the equalization equation.
Since endogeneity may go undetected due to weak instruments, James H. Stock and
Motohiro Yogo (2005) provide a rule-of-thumb test for weak instruments and critical values for
this statistic. The null hypothesis for this test is that all IVs are sufficiently correlated with the
potentially endogenous variables that the small sample bias of 2SLS is less than the bias of FE.
Using the Cragg-Donald Wald F values, the weak identification tests show that the bias of the IV
estimate of the growth equation is less than 5 percent that of the OLS estimates. In other words,
the set of instruments are sufficiently, strongly correlated with the potentially endogenous
variables that the small sample bias of 2SLS is small relative to that of OLS.
Finally, John D. Sargan (1958), Lars Peter Hansen (1982), and Lars Peter Hansen and
Kenneth J. Singleton (1982) show in a GMM context that when the number of moment
conditions exceeds the number of parameters to be estimated, a chi-square test can be used to
determine whether the instruments are orthogonal to the error term. The values of the Sargan
statistics for the test of over identifying restrictions in the growth and equalization equations are
58
5.04 and 0.07, respectively. These values fail to reject the null hypothesis that the instruments are
exogenous, providing further evidence that the instruments are valid.
To further test the robustness of these results, we use our data to conduct an “extreme-
bounds analysis” (EBA) for the growth equation, based on the work of Edward Leamer (1978,
1983, and 1985), Ross Levine and David Renelt (1992), and Tao Zhang and Heng-Fu Zou (1998).
In our base scenario, we include the variables of primary interest (I), specifically fiscal
decentralization (FD), squared fiscal decentralization and horizontal fiscal equalization (E), and
the main model variables (M), including capital (K), labor (L) as approximated by natural
logarithm of state population, central revenue share in GDP (CT), and state revenue share in
GSDP (ST). We proceed by estimating thousands of variations of the base scenario described
above, by adding every combination of the remaining 28 variables in sets including 0 to four (4)
optional variables (O). Then, we identify the control set of optional variables that give us
minimum and maximum values of the estimated coefficients of each of the primary variables of
interests and the main variables.
Table 16 summarizes the results of our extreme bounds analysis. Among the set of
interesting (I) and main (M) variables, only the estimated coefficients of central revenue share in
GDP (CT) are same signed and more than two standard deviations from zero for all
combinations up to 4 of the optional variables. Therefore, CT is the only robust variable among
all the growth determinants considered in this study. The lower bound estimate for fiscal
decentralization (FD) is negative while the base and upper bound values are positive. All the
three estimates for FD-squared are positive but the base estimate is statistically insignificant at
conventional levels. The three estimates for equalization (E) are negative but the upper and base
bounds are statistically insignificant at conventional levels. Therefore, these two variables are
59
fragile, as are the other main model variables. Based on these findings, we are somewhat assured
that it is appropriate to include fiscal decentralization, equalization, capital, labor, and central
revenue share in GDP (CT) in the growth equation.
Conclusions
We do not find evidence in support of the hypothesized trade-off between economic
growth and horizontal fiscal equalization. Instead, we find that growth has a positive effect of on
horizontal fiscal equalization. We also find that fiscal decentralization has a non-linear effect on
growth. In the 0 to 21 range, the effect is negative and for values greater than 21 the effect is
positive. In our sample, however, the index of fiscal decentralization is substantially less than 21,
meaning that fiscal decentralization is having a negative effect on growth in India. Surprisingly,
we find no evidence to support claims that the economic reform program enacted in India in the
early 1990s is having any effect on growth or horizontal fiscal equalization.
60
Chapter IV: Conclusions
In this chapter, we try to pull together the results from these two case studies of China
and India’s experience with fiscal decentralization and horizontal fiscal equalization. China and
India’s growth experiences are extremely important if for no other reason that the number of
people effected by their economic policies. Inequality, in terms of either per capita regional
income or per capita regional expenditures, still remains substantial in both China and India.
Based on the 2008 United Nations Human Development Report, about 16 percent of the
population in China and 42 percent in India are still living under 1.25 dollars a day in 2006, in
purchasing-power parity terms. Economic growth with inequality in both countries has displaced
millions of rural and urban poor from their land and livelihoods. The underbelly of this growth
experience is unprecedented level of regional disparities, unrest, and worsening human well-
being in rural areas, at least in the modern era.
Horizontal fiscal equalization is playing a role in achieving a more uniform level of
public-service provision nationwide, at least in terms of per capita expenditures. We do not find
any evidence in either country to support the hypothesized trade-off between horizontal fiscal
equalization and economic growth. In fact, we find evidence that economic growth is having a
weak positive effect of economic growth, at least in the case of India.
That our results for China differ from those of Qiao et al. (2008) may have several
potential explanations. First, in both cases, we fully account for the dual fiscal accounts, namely
budget and extra-budget accounts for China. In the case of India, we account for the revenue and
capital accounts. Second, we extend the study period to 2005, while Qiao et al. (2008) only cover
the period through 1998. The longer study period in our study better captures the effects of the
1994 TSS reform in China. Our data also straddle the 1991 economic reforms in India. Third, we
61
address the potential endogeneity of fiscal decentralization (and its second moment), equalization,
and growth. Fourth, in the case of China, we use a more complete control set, such as openness
to international trade, foreign direct investment, and the rate of inflation in the growth equation
and the share of missing women in the equalization equation. Fifth, our data are from the China
Data Center at the University of Michigan, which is a neutral third-party and therefore may be
more reliable than the data used in Qiao at al. (2008). Finally, this study examines both China
and India, and the results are generally consistent in both cases. This perhaps gives us more
confidence in concluding that there is no trade-off between equalization (E) and growth (G). Our
findings also stress the importance of accounting for extra-budget accounts in China, as Qiao et
al. (2008), Christine Wong (1998), and others stress in their writings on China.
The implications of our findings are straightforward. First of all, fiscal decentralization
has different effects in different countries, possibly because the details matter and the details
differ. Therefore, cross country growth regressions with a single scalar measure of fiscal
decentralization are undoubtedly highly problematic. For the case of India, we find substantial
evidence that economic growth contributes to horizontal fiscal equalization. Furthermore, the
policy of interregional fiscal equalization does not appear to hurt growth.
In both countries, fiscal decentralization has a non-linear effect on growth. The
relationship between fiscal decentralization and growth for China and India is illustrated in
Figure 7. Several features of this graph are noteworthy. The minimum point (FD = 10) and the
point (FD = 20, 21) where fiscal decentralization changes from having a negative to a positive
effect on growth are remarkably similar for these two countries. Finally, fiscal decentralization
seems to be having a greater impact on growth in India than in China. This may reflect the fact
62
that in India fiscal decentralization has not gone beyond the state level. Whereas, in China, the
third-tier of government or local government plays an important role in development policy.
The economic reforms in China and India are delivering tremendous economic benefits to
the people of both countries. China’s 1994 Tax Sharing System reform not only is increasing
economic growth but also equalization of per capita provincial expenditures. India’s policies of
in-depth marketization and liberalization are increasing the economic growth rate of the country,
but the effect on equalization of per capita expenditures is not as favorable as in the case of
China. This finding is somewhat surprising that India is a democracy, and China is not. This may
point to the problem with patronage politics in ethnically fragmented countries like India. But,
that is a story for another day.
63
Appendix A
Table 1: China’s Statutory Revenue Assignments in 1993 and 2007
Assignment
of functions Main Revenues
1993 2007
central sub-national central sub-national
Exclusively
assigned to
the central
government
Customs duties 100 0 100 0
Value added taxes and excise taxes on imports
100 0 100 0
Tax reimbursements for export of foreign trade businesses
100 0 100 0
Exclusively
assigned to
sub-national
governments
Urban maintenance & construction taxes (other than from railways corporations, banks, & non-bank financial institutions like insurance businesses)
0 100 0 100
Vehicle and vessel sales, use taxes 0 100 0 100
Profit remittances from sub-national SOEs 0 100 0 100
Real estate taxes in urban and township areas and land use-related taxes
0 100 0 100
Stamp taxes other than those on security 0 100 0 100
Deed taxes 0 100 0 100
Fixed asset investment regulation taxes (suspended in 2000)
0 100 0 100
Banquet taxes (mostly abolished in 2002) 0 100 n.a. n.a.
Agricultural taxes (abolished in 2005) 0 100 n.a. n.a.
Slaughter taxes (abolished in 2006) 0 100 n.a. n.a.
Shared taxes
Value Added Taxes (VAT) 12 88 75 25
Product / Excise taxes 57 43 100 0 Corporate Income Taxes (CIT) 78 22 64 36 Personal Income Taxes (PIT) 0 100 60 40 Business taxes 8 92 3 97 Stamp taxes on security transaction 50 50 97 3 Agricultural sector taxes 8 92 97 3 Other taxes n.a. n.a. 20 80
Tax revenues 21 79 58 42 Non-tax revenues 98 2 24 76 Total revenues 33 67 54 46
Sources: Document No. 85 about 1994 Tax Sharing System Reform by the State Council, other State Council documents since 1994, Finance Yearbook of China 1994, 2008 and reports from Ministry of Finance. Notes: 1) CIT in 1993 refers to SOEs only. 2) The product taxes were replaced by excise taxes/consumption tax in 1994. 3) Taxes on the agricultural sector include husbandry tax, contract tax, special agricultural product taxes, arable land use tax and other agriculture-related taxes and charges.
64
Table 2: China’s Statutory Expenditure Assignments in 1993 and 2007
Assignment of
function Main Responsibilities
1993 2007
central sub-
national central
sub-
national
Exclusive
responsibility of the
central government
National defense 100 0 100 0
Foreign affairs, aid and debt service 100 0 100 0
International trade policies 100 0 100 0
National fiscal and monetary policies 100 0 100 0
National infrastructure 100 0 100 0
Operation of the central government and central judicial organs
100 0 100 0
National projects 100 0 100 0
Technical renovation, research and development by central SOEs
100 0 100 0
Macroeconomic control, coordination among regions, redistribution like transfers among regions
100 0 100 0
Exclusive
responsibility of sub-
national governments
Local parks, recreation, fire safety and other local public services
0 100 0 100
Sub-national projects, infrastructure and housing
0 100 0 100
Technical renovation, research and development by sub-national SOEs
0 100 0 100
Operation of sub-national governments and sub-national judicial organs
0 100 0 100
Sub-national water, power, sewage, waste disposal and welfare
0 100 0 100
Concurrent
responsibility
Expenses on economic development such as capital investment and accumulation
69 31 69 31
Education and culture 11 89 7 93
Public health and sanitation 49 51 2 98
Science and technology 33 67 52 48
Social security such as pensions and unemployment insurance
1 99 6 94
Subsidies on agriculture and others 10 90 9 91
Public security 25 75 17 83
Expenditures on industry, transportation and commerce including environmental protection, urban maintenance and construction
32 68 34 66
Administration expenses 8 92 25 75
Interest payments on debts 100 0 94 6
Sources: the Constitution of People’s Republic of China in 1982 and amendments through 2004, Finance Yearbook of China 1994 and 2008. Notes: Capital construction in 2007 is calculated by the author based on reports from the Ministry of Finance in 2008 due to the expenditure accounting change in 2007.
65
Table 3: Variable Definitions for the China Regressions
D Dependent variables
E The degree of horizontal fiscal equalization measured by the distance between provincial budgetary expenditure per capita and the mean value of provincial budgetary expenditures per capita, normalized by the mean value.
G (1) The real growth rate of provincial income.
G (2) The rate of growth in nominal Gross Regional Product (GRP) per Capita.
G (3) The rate of growth in nominal GRP. G (4) The rate of growth in real GRP per Capita.
I Explanatory variables of interest
FD (1) The degree of fiscal decentralization as measured by the ratio of provincial budgetary expenditures to central budgetary expenditures in per capita terms.
FD (2) The degree of fiscal decentralization as measured by the ratio of consolidated provincial expenditures to consolidated central expenditures in per capita terms.
FD (3) The degree of fiscal decentralization measured by the marginal retention rate of provincial budgetary revenues.
FD (4) The degree of fiscal decentralization as measured by the share of provincial budgetary expenditures in the sum of provincial budgetary expenditures and central budgetary expenditures in per capita terms.
FDSQR The square of the degree of fiscal decentralization.
M Main explanatory variables
CT Central budget revenues as a share of Gross Domestic Product. K (1) The change in the stock of capital as measured by the ratio of fixed assets investment to provincial income.
K (2) The change in the stock of capital as measured by the growth rate of fixed assets investment.
L (1) The growth rate of provincial labor force. L (2) The growth rate of total population.
PT The average provincial tax rate as measured by provincial tax revenues as a share of provincial income.
PTSQR The square of the provincial tax rate to control for possible nonlinear effects.
Z Optional explanatory variables of interest
AGRS Agricultural share in GRP.
CADM The share of central budgetary administration expenditures in total central budgetary expenditures. CATD Catastrophe dummy for natural disasters like flood, earthquake, etc.
CDEV The share of central budgetary development expenditures in total central budgetary expenditures (development expenses include capital construction, enterprise upgrading, technical R&D and subsidies to agricultural sector).
CDF9 The share of central budgetary defense expenditures in total central budgetary expenditures.
CHUM The share of central budgetary human capital expenditures in total central budgetary expenditures (human capital expenditure includes culture, education, healthcare and science).
F The degree of openness measured by the share of total exports plus imports in provincial income.
FISCAP Fiscal capacity as measured by a 3-year moving average of real GRP per capita.
FPMP The price of farm products relative to the price of non-farm products as measured by the ratio of the real procurement price index of farm products to the real price index of manufactured goods in rural areas.
GOVS Government size measured by the share of consolidated expenditures in GRP
HRS A measure of the household responsibility system as measured by the percentage of production teams in rural areas that adopted the system.
L9FDI Log of foreign direct investment per capita.
L9POP Log of provincial total population.
MWS The share of missing women in total population based on methods by A. Coale and J. Banister (1994). 9SOESH The share of output value of non-State Owned Enterprises in the total value of industrial output.
9T The national tax rate defined as the share of national budgetary revenues in GDP.
PADM The share of provincial budgetary administration expenditures in total provincial budgetary expenditures.
PDEV The share of provincial budgetary development expenditures in total provincial budgetary expenditures.
PHUM The share of provincial budgetary human capital expenditures in total provincial budgetary expenditures. POPSHR The rate of growth in the rural population.
PSSS The percentage of graduates in primary schools entering into secondary schools
PURB The share of provincial budgetary urban maintenance expenditures in total provincial budgetary expenditures. R Inflation rate as measured by the percentage change in the general retail price index.
TSS Fiscal regime reform dummy, “0” for years before 1994 and “1”, otherwise.
XBGT The ratio of extra-budgetary expenditures to budgetary expenditures.
66
Table 4: Summary of Previous Findings for China
Variable
9ame
Zhang and Zou (1998)1 Lin and Liu (2000)1 Ding (2007)2 Q-M-X (2008)3
model (1)
1980-1992 model (2) 1987-1993 1970-1993 1994-2002
Growth
equation
1985-1998
Equalization
equation
1985-1998
Variables
of
interest: I
G G (1) G (1) G (2) G (3) G (3) -0.003 (-1.68) G (3)
E -41.598 (-4.92)
FD (1) -0.011 (-2.743)
-2.504 (-2.765)
FD (2) 0.217 (3.030)
FD (3) 0.0362 (2.703)
FD (4) 367.137 (3.23)
-1.687 (-5.27)
FDSQR (4)
-368.043
(-3.85)
Main
variables:
M
K (1) -0.055 (0.581)
K (2) 0.0478 (3.819)
0.093 (8.460)
0.125 (4.95)
L (1) 0.088 (0.221)
0.295 (0.685)
1.323 (7.13)
L (2) -0.209 (-1.612)
CT 0.38
(0.841)
PT -0.204 (-1.525)
-1.932 (-4.16)
-0.024 (-1.95)
PTSQR 0.05
(4.43) 0.001 (3.12)
Optional
variables:
Z
NT -14.699 (-2.514)
F 0.068 (0.750) 0.229 (5.050) R -0.120 (-0.737)
CADM 9.216 (3.283)
CDEV 1.681 (2.311)
CDFN -7.795 (-3.352)
CHUM -0.377 (-0.442)
PADM -0.332 (-1.813)
PDEV -0.254 (-2.006)
PHUM 0.304 (1.65)
PURB 0.155 (0.413)
HRS 0.0372 (1.768) FISCAP -0.144 (-5.308)
NSOESH 0.142 (3.163) FPMP 0.0107 (1.158)
POPSHR 0.0446 (0.630) XBGT 0.321 (4.99)
Obs. 308 125 534 227 392 392
Degree of
freedom 304 111 526 224 378 382
Adj. R2 0.039 0.688 0.52 0.375 0.49 0.24
Methods fixed effects random effects fixed effects fixed effects 2SLS fixed
effects 2SLS fixed
effects
Notes: T-statistics are provided in parentheses. 1Zhang and Zou (1998) and Lin and Liu (2000) include 28 provinces; exclude Chongqing, Tibet and Hainan. 2Ding (2007) includes 30 provinces; excludes Chongqing. 3Baoyun Qiao, Jorge Martinez-Vazquez and Yongsheng Xu (2008) include 30 provinces; excludes Tibet; Chongqing data are combined with Sichuan’s; and Hainan data are combined with Guangdong’s.
67
Table 5: Summary Statistics for the China Sample (1985-2006)
Variable Observation
9umber
Mean
(Standard
Deviation)
Coefficient
of
Variation
(percent)
Minimum Maximum
G (growth rate in real
GRP per capita) 630
9.45 (4.11)
43.49 -3.70 38.10
E (horizontal fiscal
equalization) 630
-0.46 (0.53)
-115.22 -3.46 0.00
FD (fiscal
decentralization) 630
3.34 (2.53)
75.75 0.86 17.85
FDSQ (FD-squared) 630 17.60
(35.78) 203.30 0.73 318.46
K (growth rate of fixed
asset investment) 630
20.46 (16.70)
81.62 -22.91 116.26
L (growth rate of
employments) 630
1.84 (3.96)
215.22 -23.18 33.17
F (degree of openness) 630 23.94
(32.38) 135.25 2.13 184.51
R (inflation rate) 630 5.86
(7.66) 130.63 -5.5 30.20
L9FDI (natural
logarithm of foreign
direct investment)
630 1.75
(2.33) 133.14 -5.21 6.18
TSS (fiscal regime
dummy, “1”for years ≥
1994, “0”otherwise)
630 0.63
(0.48) 76.19 0.00 1.00
AGRS (agriculture
share) 630
21.35 (10.07)
47.17 0.88 50.90
PT (provincial tax rate) 630 14.03 (6.76)
48.18 5.33 53.73
PTSQ (PT-squared) 630 242.36
(282.27) 116.47 28.38 2886.63
XBGT (extra-budget
factor) 630
0.44 (0.30)
68.18 0.01 1.58
CDEV (central
government development
spending)
630 45.62
(12.11) 26.55 29.01 68.03
MWS (missing women
share) 630
0.04 (0.01)
18.2 0.01 0.06
GOVS (government size) 630 0.20
(0.10) 50.00 0.08 0.83
L9POP (natural
logarithm of population) 630
17.19 (0.90)
5.23 14.57 18.40
CATD (catastrophe
dummy) 630
0.34 (0.47)
138.23 0 1
Data Source: 1) China Data Center of the University of Michigan and the Ministry of Finance; 2) Catastrophe dummy for each year, Tian’anmen Square incident dummy and Asian Financial crisis dummy are not reported. China flood data is retrieved in April 2009 from http://www.chinawater.net.cn/flood, level B and above with “1”, otherwise, “0”; other catastrophe data is retrieved in April 2009 from http://zzys.agri.gov.cn.
68
Table 6: Regression Results for China (1985-2006 provincial data)
Growth Equation
Dependent variable: G
(Growth rate in real GRP per capita)
Equalization Equation
Dependent variable: E
(Horizontal Fiscal Equalization)
Independent variables Fixed Effects 2SLS
Fixed Effects 2-Step GMM Fixed Effects
2SLS
Fixed Effects
2-Step
GMM
G (per capita real GRP
growth rate)
0.002 (0.002)
0.004 (0.006)
0.004 (0.005)
E (horizontal fiscal
equalization)
1.01 (0.98)
2.28 (1.63)
2.28 (1.58)
FD (fiscal decentralization) -1.48***
(0.36) -3.79***
(0.74) -3.79***
(0.72) 0.02
(0.014) 0.03
(0.03) 0.03
(0.03)
FDSQ (squared FD) 0.07*** (0.02)
0.19*** (0.04)
0.19*** (0.04)
-0.007*** (0.0007)
-0.01*** (0.001)
-0.01*** (0.001)
K (growth rate of fixed
asset investment)
0.11*** (0.01)
0.11*** (0.01)
0.11*** (0.01)
-0.0002 (0.0004)
-0.001 (0.001)
-0.001 (0.001)
F (openness) 0.01
(0.01) 0.02* (0.01)
0.02* (0.01)
0.0008** (0.0004)
0.001*** (0.0004)
0.001*** (0.0004)
AGRS (agriculture share) -0.22***
(0.04) -0.31***
(0.05) -0.31* (0.05)
-0.005** (0.002)
-0.004* (0.0025)
-0.004* (0.0024)
TSS (fiscal regime dummy
since 94)
2.19** (0.86)
2.92*** (0.92)
2.92*** (0.89)
0.05 (0.04)
0.09** (0.045)
0.09** (0.044)
PT (provincial tax rate) -0.08 (0.12)
-0.06 (0.13)
-0.06 (0.13)
0.03*** (0.005)
0.04*** (0.006)
0.04*** (0.006)
PTSQ (squared PT) -0.001 (0.002)
-0.0002 (0.003)
-0.0002 (0.002)
-0.001*** (0.0001)
-0.001*** (0.0001)
-0.001*** (0.0001)
CDEV (central
government development
spending)
0.13*** (0.04)
0.12*** (0.04)
0.12*** (0.04)
0.006*** (0.002)
0.008*** (0.002)
0.008*** (0.002)
GOVS (government size) 11.15** (4.47)
12.88** (5.19)
12.88*** (5.03)
-1.20*** (0.18)
-0.74*** (0.21)
-0.74*** (0.20)
L (growth rate of
employments)
0.02 (0.03)
0.03 (0.03)
0.03 (0.03)
R (inflation rate) 0.08*** (0.03)
0.13*** (0.03)
0.13*** (0.03)
L9FDI (foreign direct
investment)
0.44*** (0.13)
0.62*** (0.14)
0.62*** (0.14)
XBGT (extra-budget
factor)
-0.17*** (0.04)
-0.27*** (0.04)
-0.27*** (0.04)
L9POP (population) 0.31** (0.13)
0.26* (0.15)
0.26* (0.14)
MWS (share of missing
women)
-3.49*** (1.22)
-2.51* (1.35)
-2.51* (1.31)
Constant 7.18*** (2.49)
13.20*** (3.23)
-5.77** (2.29)
-5.26** (2.58)
9umber of observations 630 630 630 630 630 630
R-squared: 0.58 0.44 0.60 0.27 0.48 0.42
Weak identification
(Stock and Yogo 2005)
8.04 The bias of
IV estimates < 20% of
OLS
13.75 The bias of
IV estimates < 5% of OLS.
Sargan statistic 10.53
Prob=0.65
0.54 Prob=0.76
1) Standard errors are provided in parentheses. 2) *** Statistically significant at 1 percent; ** significant at 5 percent; and * significant at 10 percent. 3) Because of the simultaneity of SEM, 2SLS fixed effects and 2-step GMM for both equations are estimated with
IVs. 4) Each year’s catastrophe dummy, Tianánmen Square incident dummy and Asian Financial crisis dummy
estimates are not reported and available upon request.
69
Table 7: Specifications Tests for the China Regressions
Hausman test:
IV regression versus
regression without IV 1 Durbin-Wu-Hausman test
2
Growth equation
Fiscal Equalization
Equation
Growth Equation Fiscal Equalization Equation
E FD Squared FD G FD Squared FD
Chi-Square 14.25 43.64 0.95 0.13 3.35 0.18 27.34 54.43
P-value 0.99 0.15 0.33 0.72 0.07 0.67 0.00 0.00
Implication
of test
Fail to reject the null
Fail to reject the null
Fail to reject exogeneity
Fail to reject exogeneity
Reject the exogeneity at
10%
Fail to reject exogeneity
Soundly reject exogeneity
Soundly reject exogeneity
Notes: 1 For the Hausman test, the null hypothesis is that the two system estimates are not statistically significantly different from each other. 2 For the Durbin-Wu-Hausman tests, the null hypothesis is that the tested variable is exogenous. *The instruments for E, FD and Squared FD in the growth equation are: XBGT, CDFN, MWS, one and two lagged-value of FD, one lagged value of E, the interactive of XBGT and MWS, the interactive of XBGT and CDFN, the interactive of MWS and CDFN, squared MWS, squared XBGT, squared CDFN, the interactive of squared XBGT and CDFN, the interactive of cubic XBGT and CDFN, the cubic CDFN, the interactive of squared MWS and squared CDFN; **The instruments for G, FD and Squared FD in the equalization equation are: one and two lagged-value of FD, one lagged-value of G, squared one lagged value of G, squared one lagged value of FD.
70
Table 8: Extreme Bounds Analysis for China
(Dependent Variable: Growth Rate of Real Per Capita GRP, 1949-2008, Fixed Effect)
Variable
type Variables Coefficient
Standard
error t value
# of
Obs. R - square Other variables
Robust
/
Fragile
I
variables
FD
high 0.65 0.31 2.09 1151 0.43 CT, PHUM, CHUM, LNFISCAP
Fragile base -0.35 0.26 -1.36 1237 0.36
low -1.09 0.29 -3.72 900 0.33 PDEV, AGRS, PSSS, CATD
Squared
FD
high 0.06 0.024 2.56 900 0.33 PDEV, AGRS, PSSS, CATD
Fragile base 0.013 0.02 0.62 1237 0.36
low 0.03 0.014 1.97 702 0.50 CT, LNFDI, PADM, CDFN
E
high N/A N/A N/A N/A N/A N/A
Fragile base -0.24 0.61 -0.39 1237 0.36
low N/A N/A N/A N/A N/A N/A
M
variables
K
high 0.15 0.007 21.68 659 0.52 LNFISCAP, NT, CT, XBGT
Robust base 0.12 0.006 21.36 1237 0.36
low 0.08 0.006 14.75 904 0.35 CT, CADM, CDFN, PSSS
L
high 0.31 0.05 5.92 1164 0.38 CDEV, PDEV, CHUM, PHUM
Robust base 0.23 0.05 4.86 1237 0.36
low 0.08 0.04 2.00 914 0.36 LNFISCAP, CDFN, AGRS, PSSS
F
high 0.03 0.009 3.24 883 0.37 LNFISCAP, PADM, AGRS, PSSS
Fragile base 0.012 0.009 1.39 1237 0.36
low 0.01 0.005 1.97 654 0.53 XBGT, LNFDI, PADM, CDEV
R
high 0.10 0.02 4.69 575 0.51 CDEV, XBGT, CHUM, PSSS
Fragile base -0.006 0.007 -0.82 1237 0.36
low -0.006 0.007 -0.82 1237 0.36
Notes: 1) Regression model form: Yit=βi Iit + βm Mit+ βz Zit +µi + eit based on Levine and
Renelt (1992), where Iit are the explanatory variables of interest, Mit are the main explanatory variables, and Zit are the other explanatory variables. The base estimates are from the regression that includes the explanatory variables of interest (I-variables) and the main explanatory variables (M-variables), which are included in every regression.
2) The robust/fragile designation follows that of Levine and Renelt (1992). if the high, base, and low estimated coefficients all have the same sign and are more than two standard deviations from zero, then the variable is said to be robust; otherwise, it is said to be fragile.
3) N/A implies that no bounds are found at the conventional 95% confidence intervals.
71
Table 9: Comparison of Results for China
Q-M-X model
The original estimates of
Q-M-X, 1985-1998 1985-1998, our data 1985-2006, our data 1949-2008, our data
Growth
equation
Equalization
equation
Growth
equation
Equalization
equation
Growth
equation
Equalization
equation
Growth
equation
Equalization
equation
Intercept -75.711**
(-2.05) 0.551** (2.08)
3.99 (67.30)
0.30 (0.30)
102.84* (56.30)
-0.13 (0.17)
16.18*** (4.00)
-0.74*** (0.08)
G (growth) -0.003 (-1.68)
0.002 (0.002)
0.002
(0.0018)
-0.001 (0.001)
E (equalization) -41.598***
(-4.92)
-5.40 (7.84)
24.16** (12.05)
11.77***
(3.60)
FD (fiscal
decentralization)
367.137*** (3.23)
-1.687*** (-5.27)
79.14 196.60)
-1.05** (0.42)
-276.77** (164.84)
-0.575** (0.27)
-22.85** (11.10)
1.61*** (0.27)
Squared FD
-368.043***
(-3.85)
-103.22 (151.60)
238.49* (128.67)
40.7*** (10.00)
-2.14*** (0.26)
9umber of
observations 392 392 364 364 588 588 1461 1461
R-squared 0.51 0.26 0.23 0.01 0.06 0.002 0.30 0.17
Notes: 1) Q-M-X model refers to the model in Baoyun Qiao, Jorge Martinez-Vazquez and Yongsheng Xu (2008); 2) All of the estimates are panel 2SLS fixed effects; 3) *** Statistically significant at 1 percent; ** significant at 5 percent; and * significant at 10 percent; 4) In parentheses are standard errors of relevant estimates except those for the original estimates of Q-M-X, whose are t-statistics; 5) The ratio of extra-budget over budget expenditures is available from 1985 to 2006; 6) Employment growth rate is a proxy for labor force growth rate in (Qiao et al. 2008); 7) Other explanatory variable estimates are not reported in this table and available upon request.
72
Table 10: India’s Statutory Expenditure Assignments
Main Responsibilities Assignments
National defense and others related to intelligence & forces of the Union
List I-Union List
Foreign affairs, foreign loans & public debts
International trade, treaties, conferences, and social order on the high seas
International civil affairs like naturalization, migration, & pilgrimages, etc.
National fiscal and monetary tools like currency, Reserve Bank of India, Post Office Savings
Bank, lotteries, banking, insurance, stock market, future market, & other derivative markets.
National infrastructures like airways, railways, national highways, national
waterways, maritime shipping & navigation, lighthouses, ports, posts &
telegraphs, telephones, wireless, broadcasting, properties of the Union
Interstate trade and commerce, Intellectual Property Rights protection
Establishment of standards of weight and measure; Regulation of goods to be exported abroad or interstate
Natural resources regulation, Cultivation, manufacture, and sale for export of opium
Coordination and standardization of higher education and research
National heritages and institutions, Union public services, All-India services, Census
Elections to Parliament and Legislatures of the states, Offices of President
Operation of the Union government, Parliament and the Union judicial organs
Audit of the accounts of the Union and the States
Interstate migration and quarantine
Jurisdiction and powers of all matters on List I except the Supreme Court; Any other matter not in List II or List III
Criminal laws & procedures about the matters not in List I & II
List III-Concurrent List for the Union and
State
Transfer of property other than agricultural land; Many domestic civil laws concerning marriage, family, and so on
Contracting other than agricultural land; Bankruptcy and insolvency, Trust and Trustees
Administration of Justice except the Supreme Court and the High Courts
Vagrancy; nomadic, and migratory tribes; National environment, animal and plants protection
Economic and social planning including family planning; Commercial and industrial monopolies, combines, and trusts
Trade unions; industrial and labor disputes; Charities and religions
Social security and social insurance, welfare of labor; Education and Legal, medical, and other professions
Interstate public health; Vital statistics; Price control; Mechanically propelled vehicles, factories, boilers, electricity,
Jurisdiction and powers of all matters on List III except the Supreme Court
Public order except that subject the control of the Union
List II-State List
Operation of State governments & local governments, the Legislatures of the states, the State judicial and correction
Public goods and services such as public health & sanitation, pilgrimages, social relief, regulation of intoxicating
burials & cremations, public libraries & museums, communications not in List I, water, land, fisheries, gas,
markets & fairs, inns, sports, entertainments, gambling, incorporations other than those in List I
Agriculture, Trade and Commerce within the State
Public debt of the State, Treasure trove
Jurisdiction and powers of all matters on List II except the Supreme Court
Summarization and compilation by author based on the Seventh, Eleventh and Twelfth Schedules of the Constitution of India and the Amendments until Dec 2007.
73
Table 11: India’s Statutory Revenue Assignments
Main tax revenues Assignments
Taxes on income other than agricultural income
List I-Union List
Duties of customs including export duties
Duties of excise on tobacco and other goods
Corporation tax
Estate duty with respect to property other than agricultural land
Taxes on the capital value of the assets, exclusive to agricultural land
Estate duty with respect to property other than agricultural land
Duties with respect to succession to property other than agricultural land Terminal taxes on goods or passengers carried by railway, sea, or air, on railway
Taxes other than stamp duties on transactions in stock exchanges and future
Rates of stamp duty with respect to bill of exchange, cheques, promissory notes,
letters of credit, policies of insurance, transfer of shares, debentures, proxies,
Taxes on the sale or purchase of newspapers and on advertisements published
Taxes on the consignments of interstate goods trade or commerce
Taxes on services
Residuary tax powers not specified in List II and III
Fees in respect of any of the matters in List I
Recovery in a State of claims with respect to taxes and other public demands List III-
Concurrent List for the Union and
State
arrears of land-revenue and sums recoverable as such arrears, arising outside
Stamp duties other than duties or fees collected by means of judicial stamps, but
including rates of stamp duty
Fees with respect to any of the matters in List III
Taxes on agricultural income
List II-State List
Duties with respect to succession to agricultural land
Estate duty with respect to agricultural land
Taxes on lands and buildings
Taxes on mineral rights
Duties of excise on the following goods manufactured or produced in the state
Taxes on the entry of goods into a local area for consumption, use or sale
Taxes on the consumption or sale of electricity
Taxes on the sale or purchase of goods except newspapers and advertisements on
Taxes on advertisements other than advertisements published in the newspapers
Taxes on goods and passengers carried by road or on inland waterways
Taxes on vehicles, boats, and animals; tolls
Taxes on professions, trades, callings, and employment
Capitation taxes
Taxes on luxuries, entertainment, and gambling
Rates of stamp duty with respect to documents other than those specified in List I
Fees in respect of any of the matters in List II
Summarization and compilation by author based on the Seventh, Eleventh and Twelfth Schedules of the Constitution of India and the Amendments until Dec 2007
74
Table 12: Variable Definitions for the India Regressions
D Dependent variables
E
The degree of horizontal fiscal equalization measured by the distance between provincial budgetary expenditure per capita and the mean value of provincial budgetary expenditures per capita, normalized by the mean value.
G Growth rate of real Gross State Domestic Product (GSDP) per capita.
I Explanatory variables of interest
FD
The degree of fiscal decentralization as measured by the ratio of consolidated state expenditures to consolidated central expenditures in per capita terms (including revenue expenditures and capital expenditures).
FDSQR The square of the degree of fiscal decentralization.
M Main explanatory variables
CT Central tax rate defined as central revenue receipts as a share of nominal Gross Domestic Product (GDP).
K Log of capital outlays per capita.
L9POP Log of state population as a proxy of labor force.
ST State tax rate defined as state revenue receipts as a share of nominal GSDP.
Z Optional explanatory variables of interest
CERE The ratio of capital expenditures over revenue expenditures.
CETE The capital expenditures as a share of consolidated expenditures.
CRTR The capital receipts as a share of consolidated receipts.
ER Economic reform dummy since 1991, “1”for years 1991 and after 1991, “0”, otherwise.
FDR The fiscal dependence rate defined as central grants share in consolidated receipts.
FSR The fiscal self-reliance rate defined as state own tax and non-tax revenues as a share of total revenue receipts.
MWS The share of missing women in total population based on hypothetical ratio by Sen (1990).
CATD Catastrophe dummy for natural disasters like flood, earthquake, etc.
GDTE The transfer dependence rate defined as gross devolution as a share of consolidated expenditures.
GLTD The net central loans as a share of state outstanding liabilities.
GOVS Government size measured by the share of consolidated expenditures in nominal GSDP.
GTR Gross fiscal deficit as a share of consolidated revenues.
IPGD The interest payment and repayment as a share of gross devolution.
LEPC Log of consolidated expenditures per capita.
LRPC Log of consolidated revenues per capita. MLTD The net market loans as a share of state outstanding liabilities.
9DD The net devolution dependence defined net devolution as a share of consolidated expenditures.
9DGD The net devolution as a share of gross devolution.
9T National tax rate defined as national revenue receipts as a share of nominal GDP.
9TA Non-tax autonomy defined as own non-tax revenues as a share of total revenue receipts.
PDG The primary deficit as a share of nominal GSDP. PDTR The primary deficit as a share of total revenue receipts.
RDG The revenue deficit as a share of nominal GSDP.
RDTR The revenue deficit as a share of total revenue receipts.
TA Tax autonomy defined as sum of own tax revenues and share in central taxes as a share of total revenue receipts.
Notes: Other optional explanatory variables that produce neither highest estimates nor lowest estimates are not reported here and can be available upon request.
75
Table 13: Summary Statistics for the India Sample (1980-2005)
Variable Observation
9umber
Mean
(Standard
Deviation)
Coefficient of
Variation
(percent)
Minimum Maximum
G (growth rate in real
GSDP per capita) 562 4.67
(13.65) 292.30 -43.70 81.67
E (horizontal fiscal
equalization) 562 -0.51 (0.55)
-107.84 -4.78 0.00
FD (fiscal
decentralization) 562 1.65
(1.48) 89.70 0.27 12.18
FDSQ (FD-squared) 562 4.93
(14.58) 295.74 0.07 148.35
K (log of capital outlays
per capita) 562 5.55
(1.27) 22.88 1.07 8.71
L9POP (log of state
population as a proxy of
labor force ) 562
16.64 (1.59)
9.55 13.00 19.15
CT (central tax rate) 562 0.09
(0.006) 6.67 0.09 0.11
ST (state tax rate) 562 0.25
(0.23) 92.00 0.07 1.95
FSR (fiscal self-reliance
rate) 562
0.50 (0.26)
52.00 0.06 0.95
FDR (fiscal dependence
rate) 562
0.29 (0.24)
82.76 0.03 1.00
MWS (missing women
share) 562
0.03 (0.01)
33.33 0.003 0.06
CATD (catastrophe
dummy) 562
0.17 (0.38)
223.53 0 1
Data Sources: 1) Directorate of Economics & Statistics of respective State Governments, and for All-India -- Central Statistical Organization; 2) Reserve Bank of India; 3) Missing women data is hypothetical and calculated by author based on Sen (1990); 4) Catastrophe dummy for each year is not reported. Natural disaster data is retrieved in
October 2009 from http://www.ndmindia.nic.in
76
Table 14: Regression Results for India (1980-2005 state-wise data)
Growth Equation
Dependent variable: G
(growth rate of real GSDP per capita)
Equalization Equation
Dependent variable: E
(Horizontal Fiscal Equalization)
Independent variables Fixed Effects 2SLS
Fixed Effects
2-Step
GMM Fixed Effects
2SLS
Fixed Effects
2-Step
GMM
G (growth rate of real
GSDP per capita)
0.0008* (0.00043)
0.008*** (0.003)
0.008*** (0.003)
E (horizontal fiscal
equalization)
7.13 (4.43)
-9.61 (11.27)
-9.61 (10.96)
FD (fiscal
decentralization)
4.93 (3.75)
-62.38*** (12.71)
-62.38*** (12.35)
-0.16*** (0.03)
0.33* (0.18)
0.33* (0.18)
FDSQ (squared FD) 0.11
(0.31) 3.05*** (1.04)
3.05*** (1.02)
-0.015*** (0.002)
-0.04*** (0.01)
-0.04*** (0.01)
L9POP (log of state
population)
-19.25*** (4.30)
-50.86*** (7.89)
-50.86*** (7.67)
0.04 (0.04)
0.44*** (0.16)
0.44*** (0.16)
ER (Economic reform
dummy, “1”for years ≧≧≧≧
1991, “0”otherwise)
-4.70** (2.21)
3.76 (3.26)
3.76 (3.17)
0.12***
(0.02)
-0.01
(0.06)
-0.01
(0.06)
K (log of capital outlay
per capita)
4.04*** (1.26)
13.25*** (2.15)
13.25*** (2.09)
CT (central tax rate) 252.18* (128.96)
-54.98 (182.53)
-54.98 (177.46)
ST (state tax rate) -23.30***
(14.28) 16.87
(33.27) 16.87
(33.34)
MWS (missing women
share)
-4.28*** (1.34)
-6.75*** (2.13)
-6.75*** (2.07)
FSR (fiscal self-reliance
rate)
-0.10 (0.13)
-0.79** (0.40)
-0.79** (0.39)
FDR (grants dependence
rate)
0.12 (0.10)
0.05 (0.15)
0.05 (0.15)
Constant 282.22***
(72.09) 859.85** (139.60)
-0.74 (0.70)
-7.69*** (2.73)
9umber of observations 562 522 522 562 502 502
R-squared: 0.002 0.003 -0.44 0.70 0.36 0.10
Weak identification
(Stock and Yogo 2005)
14.55 The bias of
IV estimates < 5% of OLS
3.04 The critical values are
not availabel.
Sargan statistic 5.04
Prob=0.17
0.07 Prob=0.80
1) Standard errors are provided in parentheses. 2) *** Statistically significant at 1 percent; ** significant at 5 percent; and * significant at 10 percent. 3) Because of the simultaneity of SEM, 2SLS fixed effects and 2-step GMM for both equations are
estimated with IVs. 4) Each year’s catastrophe dummy estimate is not reported and available upon request.
77
Table 15: Specifications Tests for the India Regressions
Hausman test:
IV regression versus
regression without IV 1 Durbin-Wu-Hausman test
2
Growth equation
Fiscal Equalization
Equation
Growth Equation Fiscal Equalization Equation
E FD Squared FD G FD Squared FD
Chi-Square 48.80 10.60 3.52 52.24 8.46 12.43 6.20 0.68
P-value 0.006 0.99 0.06 0.00 0.00 0.00 0.01 0.41
Implication
of test
Soundly reject the
null
Fail to reject the null
Reject exogeneity at
10%
Soundly reject exogeneity
Soundly reject exogeneity
Soundly reject exogeneity
Soundly reject exogeneity
Fail to reject exogeneity
Notes: 1 For the Hausman test, the null hypothesis is that the two system estimates are not statistically significantly different from each other. 2 For the Durbin-Wu-Hausman tests, the null hypothesis is that the tested variable is exogenous. *The instruments for E, FD and Squared FD in the growth equation are: MWS, one lagged-value of FD, one lagged value of E, squared one lagged value of E, cubic term of one lagged value of E, the interactive between one lagged value of E and one lagged value of FD; **The instruments for G, FD and Squared FD in the equalization equation are: one lagged-value of FD, one lagged-value of G, squared one lagged value of G, squared one lagged value of FD.
78
Table 16: Extreme Bounds Analysis for India
(Dependent Variable: Growth Rate of Real Per Capita GSDP, 1980-2005, Fixed Effects)
Variable
type Variables Coefficient
Standard
error
t-
value
9umber of
Observations
R-
square Other variables
Robust
/
Fragile
I -
variables
FD
high 6.09 2.80 2.18 508 0.06 FSR, NT, GOVS, IPGD
Fragile base 0.84 2.10 0.40 596 0.02
low -17.03 4.82 -3.53 499 0.07 TA, MLTD, PDG, LEPC
Squared FD
high 1.09 0.29 3.78 580 0.07 FSR, RDG, LEPC, GLTD
Fragile base 0.03 0.21 0.14 596 0.02
low 0.45 0.23 1.97 596 0.06 CRTR, LRPC, CETE, NT
E
high -0.15 2.62 -0.06 596 0.02
Fragile base -0.15 2.62 -0.06 596 0.02
low -15.66 4.10 -3.82 499 0.07 TA, MLTD, PDG, LEPC
M -
variables
K
high N/A N/A N/A N/A N/A N/A
Fragile base 0.23 1.33 0.17 596 0.02
low N/A N/A N/A N/A N/A N/A
L9POP
high 2.21 1.12 1.97 582 0.07 FSR, CRTR, LRPC, DR
Fragile base -0.24 0.92 -0.27 596 0.02
low -0.24 0.92 -0.27 596 0.02
CT
high 1476.13 353.88 4.17 500 0.06 PDTR, IPGD, GLTD, NT
Fragile base 236.76 101.19 2.34 596 0.02
low 197.86 100.13 1.98 596 0.06 FSR, RDTR, KG, LEPC
ST
high 44.87 20.53 2.19 549 0.04 CETE, GOVS, MLTD, NT
Fragile base -9.30 6.98 -1.33 596 0.02
low -52.90 12.39 -4.27 596 0.08 FSR, CRTR, LRPC, RDG
Notes: 1) Regression model form: Yit=βi Iit + βm Mit+ βz Zit +µi + eit based on Levine and Renelt
(1992), where Iit are the explanatory variables of interest, Mit are the main explanatory variables, and Zit are the other explanatory variables. The base estimates are from the regression that includes the explanatory variables of interest (I-variables) and the main explanatory variables (M-variables), which are included in every regression.
2) The robust/fragile designation follows that of Levine and Renelt (1992). if the high, base, and low estimated coefficients all have the same sign and are more than two standard deviations from zero, then the variable is said to be robust; otherwise, it is said to be fragile.
3) N/A implies that no bounds are found at the conventional 95% confidence intervals.
79
Appendix B
Figure 1: Trends in the Coefficients of Variation
for Provincial Expenditures Per Capita in China, 1985-2006
Notes: 1) Data source: National Statistics of Bureau and Ministry of Finance, China 2) The transfer data is not available before 1995 since the transfer system is
introduced by TSS in 1994; 3) The distance between the square-dotted line and the diamond-dotted
line is the equalization effect of transfers; 4) The distance between the square-dotted line and the triangle-dotted
line is the equalization effect of fiscal deficits/surplus.
80
Figure 2: Shares of Total Revenues and Expenditures in GDP and the
Share of Central Revenues in Total Revenues, for China (1985-2006)
Data Source: National Statistics of Bureau, China
81
Figure 3: The 9on-Linear Effect of Fiscal Decentralization on Growth for China
82
Figure 4: Trends in the Coefficients of Variation
for State Expenditures Per Capita in India, 1980-2005
Data Source: Reserve Bank of India.
83
Figure 5: Center and State Revenue and Expenditure Shares
in India, 1980-2005
Data Source: Budget documents of the Government of India and the State Governments.
84
Figure 6: The 9on-Linear Effect of Fiscal Decentralization on Growth in India
85
Figure 7: The 9on-Linear Effect of Fiscal Decentralization on Growth
for China and India
Notes: 1) The red curve represents China.
2) The blue curve represents India.
86
References
Bahl, R., Heredia-Ortiz, E., Martinez-Vazquez, J., & Rider, M. (2005a). India: Fiscal condition
of the states, international experience, and options for reform: Volume 1. Andrew Young
School of Policy Studies, Georgia State University, International Studies Program
Working Paper Series.
Bahl, R., Heredia-Ortiz, E., Martinez-Vazquez, J., & Rider, M. (2005b). India: Fiscal Condition
of the States, International Experience,and Options for Reform: Volume 2. Andrew
Young School of Policy Studies, Georgia State University, International Studies Program
Working Paper Series.
Bahl, R. W. (1999). Fiscal policy in China: Taxation and intergovernmental fiscal relations: San
Francisco: The 1990 Institute.
Bahl, R. W. (2005). Fiscal Decentralization 101. Class lecture. Andrew Young School of Policy
Studies, Georgia State University.
Bahl, R. W., & Linn, J. F. (1992). Urban public finance in developing countries. Oxford, New
York, Toronto, and Melbourne: Oxford University Press for the World Bank.
Barro, R. J. (1990). Government spending in a simple model of endogenous growth. Journal of
Political Economy, 98(5), 103–126.
Barro, R. J., & Sala-i-Martin, X. (1995). Economic growth. New York: McGraw-Hill.
Besley, T., & Coate, S. (2003). Centralized versus decentralized provision of local public goods:
A political economy approach. Journal of Public Economics, 87(12), 2611–2637.
Bird, R. M. (1993). Threading the fiscal labyrinth: Some issues in fiscal decentralization.
�ational Tax Journal, 46(2), 207–227.
Boadway, R. (2006). Handbook of fiscal federalism. Cheltenham, UK: Edward Elgar.
87
Boex, J. & Martinez-Vazquez, J. (2006). Local government financial reform in developing
countries: the case of Tanzania. Palgrave Macmillan
Bound, J., Jaeger, D. A., & Baker, R. M. (1996). Problems with instrumental variables
estimation when the correlation between the instruments and the endogenous explanatory
variables is weak. Journal of the American Statistical Association, 90(430), 443–450.
Bowman, K. J. (2007). Knowledge stocks by distance to frontier: linking low education
inequality to high growth in developing countries. Journal of Asian Economics, 18(4),
613–635.
Buse, A. (1992). The bias of instrumental variable estimators. Econometrica, 60(1), 173–180.
Chen, S., & Ravallion, M. (2008). The developing world is poorer than we thought, but no less
successful in the fight against poverty. from http://econ.worldbank.org/docsearch
Chow, G. (2006). Are Chinese Official Statistics Reliable? CESifo Economic Studies, Oxford
University Press, 52(2), 396–414.
Coale, A. J., & Banister, J. (1994). Five decades of missing females in China Demography, 31(3),
459–479.
Davoodi, H., & Zou, H.-f. (1998). Fiscal decentralization and economic growth: a cross-country
study. Journal of Urban Economics, 43(2), 244–257.
Ding, Y. (2005). Income inequality and fiscal decentralization in China, 1994–2002 Paper
presented at the An International Symposium on Fiscal Policy, Monetary Policy and
Economic Growth. from http://www.wise.xmu.edu.cn/Seminar/index.htm#
Ding, Y. (2007). Fiscal decentralization and economic growth in China, 1994–2002. Journal of
Chinese Economic and Business Studies, 5(3), 243–260.
88
Dollar, D. and B. Hofman (2008). Intergovernmental Fiscal Reforms, Expenditure Assignment,
and Governance. Public Finance in China: Reform and Growth for a Harmonious Society.
J. Lou and S. Wang. Washington D.C., The World Bank: 39-51.
Durbin, J. (1954). Errors in variables. International Statistical Review 22, 23–32.
Ebel, R. D., & Yilmaz, S. (2002). On the Measurement and Impact of Fiscal Decentralization.
Policy Research Working Paper 2809 of World Bank Institute.
Fair, R. C. (1970). The estimation of simultaneous equation models with lagged endogenous
variables and first order serially correlated errors. Econometrica, 38(3), 507–516.
Green, R. K. (2007a). Comments: GSE loan purchases, the FHA, and housing outcomes in
targeted, low-income neighborhoods. Brookings-Wharton Papers on Urban Affairs
(2007), 241–252.
Green, R. K. (2007). Comments: GSE loan purchases, the FHA, and housing outcomes in
targeted, low-income neighborhoods. Brookings-Wharton Papers on Urban Affairs, 241–
252.
Guess, G., Loehr, W., & Martinez-Vazquez, J. (1997). Fiscal decentralization: A methodology
for case studies.
Hansen, L. P. (1982). Large sample properties of generalized method of moments estimators.
Econometrica, 50, 1029–1054.
Hansen, L. P., & Singleton, K. J. (1982). Generalized instrumental variables estimation of
nonlinear rational expectations models. Econometrica, 50(5), 1269–1286.
Hausman, J. A. (1978). Specification tests in econometrics. Econometrica (46), 1251–1271.
Hayek, F. A. (1945). The use of knowledge in society. The American Economic Review, 35 (4),
519–530.
89
Heredia-Ortiz, E., & Rider, M. (2005). India's Intergovernmental Transfer System and the Fiscal
Condition of the States.
Heredia-Ortiz, E., Rider, M., & Diamond, J. (2006). India's Intergovernmental Transfer System
and the Fiscal Condition of the States Proceedings: �inety-eighth Annual Conference on
Taxation, Miami, Florida, �ovember 17–19, 2005, and Minutes of the Annual Meeting of
the �ational Tax Association, Thursday, �ovember 17, 2005 (pp. 362–371): Washington,
D.C.: National Tax Association.
Hofman, B. and S. C. Guerra (2005). Fiscal Disparities in East Asia: How Large and Do They
Matter? East Asia Decentralizes: Making Local Government Work in East Asia.
Washington D.C., The World Bank: 67-84.
Jenkins, R. (2006). Democracy, development and India's struggle against corruption. Public
Policy Research, 13(3), 147–155.
Jiang, Q., Feldman, M. W., & Jin, X. (2005). Estimation of the number of missing females in
China: 1900–2000. Chinese Journal of Population Science(4), 2–12.
Jiang, X. (2004). FDI in China: Contributions to Growth, Restructuring, and Competitiveness.
Beijing: Nova Publishers.
Jiwei, L., & Ahmad, E. (1997). Constraints in Reforming the Transfer System in China
Financing decentralized expenditures: An international comparison of grants (pp. 349–
360): Studies in Fiscal Federalism and State-Local Finance. Cheltenham, U.K., and Lyme,
N.H.: Elgar.
Johansson, P.-O. (1991). An introduction to modern welfare economics: Cambridge University
Press.
90
Kee, W. S. (1977). Fiscal decentralization and economic development. Public Finance Quarterly,
5 (1), 79–97.
Kezdi, G. (2003). Robust Standard Error Estimation in Fixed-Effects Panel Models Working
Paper, Central European University - Department of Economics.
Leamer, E. (1978). Specification searches: Ad hoc inference from non-experimental data. New
York: Wiley.
Leamer, E. (1983). Let's take the con out of econometrics. American Economic Review (73), 31–
43.
Leamer, E. (1985). Sensitivity analyses would help. American Economic Review (75), 308–313.
Levine, R., & Renelt, D. (1992). A sensitivity analysis of cross-country growth regressions.
American Economic Review, 8 (4).
Lin, J. Y., & Liu, Z. (2000). Fiscal decentralization and economic growth in China. Economic
Development and Cultural Change, 49(1), 1–21.
Litvack, J., & Seddon, J. (1999). Decentralization briefing notes. World Bank Institute (WBI)
Working Papers, no. 37142. The World Bank: Washington, D.C.
Lou, J. (2008). Public finance in China : reform and growth for a harmonious society
Washington, D.C.: World Bank.
Mankiw, G., Romer, D., & Weil, D. N. (1992). A contribution to the empirics of economic
growth. Quarterly Journal of Economics 107 (2), 407–437.
Martinez-Vazquez, J., & McNab, R. (1997). Fiscal Decentralization, Economic Growth, and
Democratic Governance (pp. 43 pages): International Studies Program, Andrew Young
School of Policy Studies, Georgia State University, International Studies Program
Working Paper Series, at AYSPS, GSU.
91
Martinez-Vazquez, J., & Rider, M. (2006). Fiscal decentralization and economic growth: a
comparative study of China and India. Indian Journal of Economics and Business, 29–46.
McLure, C. E., Jr. (1995). The Dangers of Decentralization: Comment. World Bank Research
Observer, 10(2), 221–226.
McMillan, M. L. (2007). Intergovernmental Transfers and Rural Local Governments. In R.
Boadway & A. Shah (Eds.), Intergovernmental Fiscal Transfers: Principles and Practice.
Washington D.C.: The World Bank.
Meloche, J.-P., Vaillancourt, F., & Yilmaz, S. (2004). Decentralization or fiscal autonomy ?
What does really matter ? effects on growth and public sector size in European transition
countries. Policy Research Working Paper Series 3254, The World Bank.
Oates, W. E. (1972). Fiscal federalism. New York: Harcourt Brace Jovanovich.
Oates, W. E. (1999). An essay on fiscal federalism. Journal of Economic Literature, 37 (3),
1120–1149.
Okun, A. (1975). Equality and Efficiency: The Big Tradeoff Washington D.C.: The Brookings
Institution.
Parker, E., & Thornton, J. (2007). Symposium on Russia and China: fiscal centralisation and
decentralisation in Russia and China. Comparative Economic Studies 49(4), 514–542.
Pommerehne, W. W. (1997). Quantitative aspects of federalism: A study of six countries. In W.
Oates (Ed.), The political economy of fiscal federalism (pp. 275–355). Lexington, MA:
Heath-Lexington.
Prasad, E., & Rumbaugh, T. (2004). China's growth and integration into the world economy.
IMF Occasional Paper, no. 232. Washington, D.C.: IMFo.
92
Pritchett, L. (2009). A review of Edward Luce's in spite of the Gods: the strange rise of modern
India. Journal of Economic Literature, 47(3), 771–780.
Prud'homme, R. (1995). The dangers of decentralization. World Bank Research Observer, 10 (2),
201–220.
Purfield, C. M. (2004). The Decentralization Dilemma in India (pp. 30 pages): International
Monetary Fund, IMF Working Papers: 04/32.
Qian, Y., & Weingast, B. R. (1997). Federalism as a commitment to reserving market incentives.
Journal of Economic Perspectives, 11 (4), 83–92.
Qiao, B., Martinez-Vazquez, J., & Xu, Y. (2008). The tradeoff between growth and equity in
decentralization policy: China's experience. Journal of Development Economics, 86 (1),
112–128.
Qiao, B., & Shah, A. (2006). Local government organization and finance in China. In A. Shah
(Ed.), Local governance in developing countries. Washington D.C.: World Bank
Publications.
Rao, M. G. (2003). Fiscal Decentralization in China and India: A Comparative Perspective. Asia-
Pacific Development Journal, 10(1), 25–45.
Rao, G., & Singh, N. (1999). The assignment of taxes and expenditures in India. Working paper,
University of California Santa Cruz, Department of Economics.
Rawski, T. G. (2001). "What is happening to China's GDP statistics?" China Economic Review
12(4): 347-354.
Sala-i-Martin, X. (1996). Regional cohesion: Evidence and theories of regional growth and
convergence. European Economic Review (40), 1325–1352.
93
Sargan, J. D. (1958). The estimation of economic relationships using instrumental variables.
Econometrica, 26(3), 393–415.
Sargan, J. D. (1961). The maximum likelihood estimation of economic relationships with
autoregressive residuals. Econometrica, 29, 414–426.
Sen, A. (1990). More than 100 million women are missing. �ew York Review of Books, 37
Shen, K., & Fu, W. (2005). Fiscal decentralization and regional economic growth in China.
Management World, 1(1), 31–39.
Shen, K., & Li, J. (2003). An empirical study on the influencing mechanism of the trade
development and economic growth in China. Economic Research Journal(5), 32–41.
Smart, M. (2007). The incentive effects of grants. In R. Boadway & A. Shah (Eds.),
Intergovernmental Fiscal Transfers Principles and Practice (pp. 203–230). Washington
D. C.: The World Bank.
Solow, R. M. (1956). A contribution to the theory of economic growth. Quarterly Journal of
Economics, 70(1), 65–94.
Spahn, P. B. (2007). Equity and efficiency aspects of interagency transfers in a multi-
government framework. In R. Boadway & A. Shah (Eds.), Intergovernmental fiscal
transfers: Principles and practice. Washington D. C.: The World Bank.
Stigler, G. J. (1957). The tenable range of functions of local government. In U. C. J. E.
Committee (Ed.), Federal expenditure policy for economic growth and stability.
Government Printing Office: Washington, D.C.
Stock, J. H., & Yogo, M. (2005). Testing for Weak Instruments in IV Regression. In D. W. K.
Andrews & J. H. Stock (Eds.), Identification and Inference for Econometric Models: A
Festschrift in Honor of Thomas Rothenberg (pp. 80–108): Cambridge University Press.
94
Swan, T. (1956). Economic growth and capital accumulation. Economic Record, 32(334–61).
Tanzi, V. (1996). Fiscal federalism and decentralization: A review of some efficiency and
macroeconomic aspects. In Annual world bank conference on development economics,
1995 (pp. 295–316). Washington, D.C.: The World Bank.
Tiebout, C. M. (1956). A pure theory of local expenditures. The Journal of Political Economy,
64 (5), 416–424.
Tobin, J. (1956). The interest-elasticity of transactions demand for cash. The Review of
Economics and Statistics, 38(3), 241–247.
Tseng, W. (1994). Economic reform in China: a new phase. Washington DC: International
Monetary Fund.
Tseng, W., Khor, H. E., Kochhar, K., Mihaljek, D., & Burton, D. (1994). Economic Reform in
China: A �ew Phase: Occasional Paper No. 114, IMF.
Wallis, J. J., & Oates, W. E. (1988). Decentralization in the public sector: An empirical study of
state and local government. In Fiscal federalism: Quantitative studies (pp. 5–28).
National Bureau of Economic Research Project Report Series. Chicago and London:
University of Chicago Press.
Wang, S. (1997). The Bottom of Decentralization (1 ed.). Beijing: China Planning Press.
Wasylenko, M. (1987). Fiscal decentralization and economic development. Public Budgeting
and Finance, 7 (4), 57–71.
Wong, C. (1998). Fiscal dualism in China: gradualist reform and the growth of off-budget
finance. In D. Brean (Ed.), Taxation in Modern China. Routledge Press, New York.
Wong, C. (2009). Global giant : is China changing the rules of the game? . New York: Palgrave
Macmillan.
95
Wooldridge, J. M. (2002). Econometric Analysis of Cross Section and Panel Data. Cambridge,
MA: MIT Press.
Wooldridge, J. M. (2003). Introductory Econometrics: A Modern Approach (2 ed.): South-
western, a division of Thomson Learning.
World Bank (2000). China: Managing Public Expenditures for Better Results.
World Bank (2001). China: Provincial Expenditure Review.
Wu, D.-M. (1974). Alternative tests of independence between stochastic regressors and
disturbances: Finite sample results. Econometrica, 42 (3), 529–546.
Zhang, B. (2006). On Measuring the Scale and Effects of Fiscal Equalization Grants in China.
Paper presented at the the Association for Chinese Economic Studies (Australia)
(ACESA) 2006.
Zhang, T., & Zou, H.-F. (1998). Fiscal decentralization, public spending, and economic growth
in China. Journal of Public Economics, 67(2), 221–240.
Zhang, Z., & Martinez-Vazquez, J. (2003a). The System of Equalization Transfers in
China.Unpublished manuscript.
Zhang, Z., & Martinez-Vazquez, J. (2003b). The System of Equalization Transfers in China.
International Studies Program Working Paper Series, at AYSPS, GSU paper0312
International Studies Program, Andrew Young School of policy Studies, Georgia State
University.
96
VITA
Yinghua Jin was born in 1973 in China. He graduated from Nanjing University of
Finance and Economics in China with a Bachelor’s degree in Economics in 1996. After
graduation, he worked in Guangdong (also Canton) province for seven years. In 2003, he came
to Georgia State University and obtained his Master’s degree in Economics in 2005. He earned
his Ph.D. in Economics from the Andrew Young School of Policy Studies, Georgia State
University, in 2009.
Jin’s research interests focus on public finance, local governance and urban economics.
His current research interests are fiscal decentralization, economic growth and fiscal equalization
in transition economies. Jin is also interested in the relationship between law and economics.
Jin can be reached at: 1322 Briarwood Rd, Apt H-10, Atlanta GA 30319
Top Related